© 2012 jian yang - connecting repositories · the alpha relaxation time ... rajarshi chakrabarti,...
TRANSCRIPT
© 2012 Jian Yang
GLASSY VISCOELASTICITY OF
DENSE SUSPENSIONS OF SOFT COLLOIDS AND
STRUCTURE & MISCIBILITY OF
SOFT FILLER POLYMER NANOCOMPOSITES
BY
JIAN YANG
DISSERTATION
Submitted in partial fulfillment of the requirements
for the degree of Doctor of Philosophy in Materials Science and Engineering
in the Graduate College of the
University of Illinois at Urbana-Champaign, 2012
Urbana, Illinois
Doctoral Committee:
Professor Kenneth S. Schweizer, Chair
Professor Steve Granick
Associate Professor Dallas R. Trinkle
Assistant Professor Charles M. Schroeder
ii
ABSTRACT
In this thesis, we studied two soft condensed matter systems. The first part of the
thesis studied the glassy viscoelasticity of dense suspensions of soft colloids and the
second part of the thesis studied the structure and miscibility of soft filler polymer
nanocomposites (PNCs).
In the first part of the thesis, we have presented the first microscopic theoretical
study of activated glassy dynamics in dense fluids of finite range soft repulsive particles
(many arm star polymers and microgel-like Hertzian spheres). The alpha relaxation time
in the activated hopping regime is a rich function of volume fraction and temperature,
including exhibiting a maximum value at ultra-high volume fraction due to a soft
jamming crossover that signals local packing disorder due to particle overlap. A kinetic
arrest diagram is constructed, and its qualitative features agree with the dynamic
crossover (MCT) analog. The isothermal dynamic fragility varies over a wide range, and
soft particles are predicted to behave as strong glasses. The highly variable dependences
of the relaxation time on temperature and volume fraction are approximately collapsed
onto two distinct master curves. We have applied NLE theory to study how particle
softness influences the elastic shear modulus, the connections between the modulus (a
short time property) and activated relaxation (a long time property), and the nonlinear
rheological effects of stress-induced yielding, shear thinning of the relaxation time and
viscosity, and stress versus shear rate flow curves of the repulsive Hertzian contact model
of soft sphere fluids.
iii
In the second part of the thesis, we have presented representative results based on
a new minimalist multi-scale model constructed for soft nanoparticle fillers. This is the
first theoretical study of the role of nanoparticle morphology associated with surface
corrugation and fluctuation at one and two particle limit in polymer melts with/without
interfacial cohesion.
Our results provide a physical basis for the unexpected ability to disperse
chemically matched crosslinked polystyrene in linear polystyrene melt. The surface
corrugation in the frozen surface regime results in a favorable entropic driving force for
mixing which competes with unfavorable depletion, resulting in a major enhancement of
nanoparticle dispersion, including a much larger spinodal solubility limit. Smooth hard
sphere behavior is recovered when the number of beads are significantly large (N=282)
and the relative corrugation size significantly reduced (from 20% to 10% of particle size).
When surface fluctuation exists, the dependence of solubility limit on fluctuation
magnitude is somewhat subtle, due to the relevance of multiple length scales. We find
fluctuation suppresses dispersion for all particles sizes (or surface curvature) studied, and
this effect is most pronounced when the monomer size is smaller than corrugation size.
The extreme coherent fluctuation model shows dramatically enhanced dependence on
fluctuation magnitude at large fluctuation magnitude and less miscibility. When
interfacial cohesion exists, we develop a model for local bead-monomer level attraction,
which explicitly connects to particle-monomer level attraction through potential mapping
strategies. By varying the interfacial attraction strength, we can still observe all three
regimes reported in prior PRISM studies of smooth HS filler PNCs. The steric
stabilization regime is not sensitive to surface fluctuation magnitude, while the contact
iv
clustering and strong bridging regimes are very sensitive to surface fluctuation magnitude.
Surface fluctuation smears out surface corrugation and reduces interfacial cohesion,
leading to stronger depletion attraction and more negative second virial coefficients.
v
To mom and dad
vi
ACKNOWLEDGEMENTS
First of all, I would like to express my sincerest gratitude to my advisor, Professor
Kenneth S. Schweizer, for all his inspiring guidance, encouragement and support
throughout my life during the graduate studies. His love for science has always inspired
me and will continue so for my future career. His physical insights over a broad range of
soft condensed matter systems always amaze me. My only regret, if there is any, during
my graduate studies, is that I could have learnt more from him. I really appreciate his
kindness, patience, and efforts to teach me about science and research. I have kept many
notes he wrote to me, as these words consistently encourage me to move forward and
gain power from the curiosity and enthusiasm for science. These years of study under his
guidance will definitely be my lifetime treasure.
I would like to thank members of my Preliminary Examination and Dissertation
Committees, Prof. Steve Granick, Prof. Dallas R. Trinkle, Prof. Charles M. Schroeder
and Prof. Kenneth S. Schweizer for their time, interest, and helpful comments.
I am also thankful to all past and present members of the Schweizer research group:
Galina Yatsenko, Arthi Jayaraman, Kang Chen, Douglas Viehman, Everett Scheer, Lisa
Hall, Mukta Tripathy, Rajarshi Chakrabarti, Rui Zhang, Daniel Sussman, Umi
Yamamoto, Debapriya Banerjee, Ryan B. Jadrich, Homin Shin, Stephen Mirigian and
Zachary E. Dell. They have provided much help in both my academic and daily life. The
conversations with them are always enjoyable. The many journal clubs we have not only
motivated my enthusiasm to research but also made research life interesting and inspiring.
vii
My special thanks go to Dr. Bo Wang and Prof. Steve Granick for their insightful
discussions and helpful suggestions from an experimental point of view. This gives me a
precious opportunity to deepen my understanding of science from a different perspective.
Finally, I express my endless gratitude to my parents, Mingquan Yang and Junhua
Xue, who consistently give me love and support.
This research was supported by the following funding sources: NSF-NSEC (RPI-
UIUC) and DOE-BES (ORNL).
viii
TABLE OF CONTENTS
Chapter 1: Introduction………………………………………………..……………...1
1. 1 Background and Motivation……………………..…………………………...1
1.2 Dissertation Overview………………………………………………………...7
1.3 References……………………………………………………………………..8
1.4 Figures………………………………………………………………………..11
Chapter 2: Glassy Dynamics and Mechanical Response in Dense Fluids of
Soft Repulsive Spheres: Methodology...……………………………….. …13 2.1 Introduction ………………………………………………………………….13
2.2 Equilibrium Liquid State Theory: Ornstein–Zernike Theory………………..17
2.3 Naïve MCT Dynamical Crossover…………………………………………...19
2.4 Quiescent NLE Theory……………………………………….………….. …19
2.5 Ultralocal Limit……………………………………………………………....21
2.6 Mechanical Response……………………………………………………...…22
2.7. References………………………………………………………………...…27
2.8 Figures……………………………………………………………………..…32
Chapter 3: Glassy Dynamics and Elasticity of Dense Fluids of Many Arm
Stars………………………………………………………………………...33 3.1 Introduction…………………………………………………………………..33
3.2 Model, Equilibrium Structure and “Soft Jamming” Crossover……………...34
3.3 Naïve MCT Kinetic Arrest Map……………………………………………..35
3.4 Activated Barrier Hopping and Dynamical Fragility………………………...36
3.5 Linear Elastic Shear Modulus and Experiment-Theory Confrontations…..…38
3.6 Ultralocal Analysis…………………………………………………………...39
3.7 Summary and Discussion………………………………………………….…40
3.8 References……………………………………………………………………41
3.9 Figures……………………………………………………………………..…44
Chapter 4: Glassy Dynamics and Mechanical Response in Dense Fluids of
Hertzian Spheres…………………………………………………………...53 4.1 Introduction…………………………………………………………………..53
4.2 Model, Equilibrium Structure and Dynamic Crossover……………………..54
4.3 Effect of Particle Softness on Dynamic Free Energy………………………..59
4.4 Mean Alpha Relaxation Time………………………………………………..62
4.5 Kinetic Arrest Map and Dynamic Fragility………………………………….66
4.6 Linear Shear Modulus………………………………………………………..68
4.7 Connections between the Elastic Modulus, Alpha Relaxation and Kinetic
Arrest……………………………………………………………………………..70
4.8 Nonlinear Rheology………………………………………………………….72
4.9 Summary and Discussion…………………………………………………….78
4.10 References…………………………………………………………………..80
4.11 Figures………………………………………………………………………84
ix
Chapter 5: Theory of the Structure and Miscibility of Soft Filler Polymer
Nanocomposites: Methodology…………………………………………...109 5.1 Introduction…………………………………………………………………109
5.2 Minimalist Model of Soft Nanoparticles and Effective Interactions based on
Small Scale Monte Carlo Simulation…………………………………………...113
5.3 Equilibrium Theory of Polymer Nanocomposites………………………….119
5.4 Combining Small Scale Monte Carlo Simulation with PRISM Theory: Three-
Step Approach…………………………………………………………………..122
5.5 References…………………………………………………………………..125
5.6 Figures………………………………………………………………………129
Chapter 6: Theory of the Structure and Miscibility of Soft Filler Polymer
Nanocomposites: Results and Discussion………………………………...132 6.1 Introduction ………………………………………………………………...132
6.2 Effective Interactions ………………………………………………………133
6.3 Athermal Systems: Pair Correlations & Thermodynamics…………………135
6.4 Role of Interfacial Cohesion……………………………………………..…146
6.5 Summary and Discussion…………………………………………………...150
6.6 References…………………………………………………………………..153
6.7 Figures……………………………………………………………………....154
Chapter 7: Multi-Scale Modeling of Soft Colloid Melts………………….195 7.1 Introduction…………………………………………………………………195
7.2 Model, Theory and Effective Volume Fraction…………………………….196
7.3 Effect of Particle Size and Softness on Mean Alpha Relaxation Time and
Kinetic Arrest…………………………………………………………………...199
7.4 Effect of Particle Size and Softness on Linear Shear Modulus…………….201
7.5 Summary and Discussion…………………………………………………...202
7.6 References…………………………………………………………………..203
7.7 Figures………………………………………………………………………205
Chapter 8: Summary and Outlook……………………………………...…210
1
CHAPTER 1
INTRODUCTION
1.1 Background and Motivation
The term “soft particle” refers to objects with tunably soft repulsive interactions
of size that range from several nanometers to greater than a micron. Soft particles can be
obtained conceptually in two ways. One is by exploiting the softness of an electrostatic or
magnetic repulsive potential. Another way is to modify the synthesis of the particles, e.g.
use many arm star polymers, crosslinked nanogels or microgels, micellar diblock
copolymers based on self-assembly, a solid particle covered with adsorbed or grafted
polymer chains, emulsion droplet, multilamellar vesicle and liposome (see Fig.1.1) [1].
Dense fluids of soft particles, or soft particle pastes, are generally deformable
objects dispersed in a solvent and in many materials applications are at large volume
fractions, often above closepacking. They form an important class of materials at the
interface between polymer solutions, granular materials, and colloids, for example, as
rheology modifiers in ceramic processing, surface coating of submicron particles [2], and
the food and pharmaceutical industries. All these materials exhibit both solid-like and
liquid-like mechanical properties, with the solid-to-liquid transition taking a variety of
forms. Such behavior is exploited industrially to formulate food or personal care products
and to process high performance materials such as films, coatings, solid inks, and
ceramics.
Compared to conventional hard sphere colloidal suspensions and glasses, soft
particles exhibit additional new and interesting structural and dynamical features. A key
2
issue in repulsive dense systems is the topological caging constraints a tagged object feels
due to its neighbors. Linear flexible polymers in solution at very high volume fraction or
in the melt can be entangled and experience the so-called tube confinement, while long
time dynamics proceeds via anisotropic reptation motion [3, 4]. On the other hand, when
hard spheres reach a volume fraction of about 0.58, they are sufficiently constrained by
neighbors (a small number not exceeding 12) that motion (on the scale of its size)
becomes arrested, forming an effective cage [5-7], a “kinetic” glass transition occurs.
Soft particles lie in between these two extremes, and a modified cage effect exists [8].
The particles interact through potentials that can be tuned over very wide range via the
chemical composition and architecture, changing from polymer coils to hard sphere
colloids. Either because of their (limited) interpenentration (many arm stars) and/or
deformability (crosslinked microgels), suspensions of such particles can reach very high
volume fractions, even beyond random close packing of hard spheres ( 0.64RCP
φ ≈ ). The
particles then develop repulsive forces of elastic origin at contact, which control the cage
elasticity and other macroscopic properties [1].
Two types of soft particles are of particular interest in this thesis. One is many
arm stars, and the other is crosslinked microgels (see Fig.1.1). Many arm star polymers
are a pristine model system for soft particles with many unique advantages. (1) Superior
structural control compared with most soft particle systems, which are not so well
characterized: polydipsersity, limited control of aggregation or grafting density, difficulty
in reproducing exactly the same composite particle, particle stability (particularly with
respect to time and to temperature). Many arm stars can be tuned from polymer coils to
almost hard spheres by increasing arm number. (2) For most soft particle systems, the
3
softness is tuned by various external parameters, e.g. temperature, pH, solvent, and
additives, while for stars, varying the number and/or size of the ‘grafted’ chains is enough.
(3) Stars have a well-established microscopic description interaction potential at the
center of mass level. Given the complexity of interactions, model systems with as simple
as possible, well-understood interactions, are highly appreciated, e.g. no charges, no
enthalpic, only entropic (excluded volume) interactions of tunable strength and spatial
range [8].
Microgel particles are cross-linked latex or collapsed polymer particles that are
swollen in a good solvent. Microgels have become one of the most popular soft matter
systems to study because they can be conveniently prepared by surfactant-free emulsion
polymerization (SFEP), and because their softness can be manipulated by varying the
amount of crosslinking or temperature. The influence of the softness of concentrated
microgel suspension dynamics has been studied in detail by Dynamic Light Scattering
(DLS) [9]. Quite remarkably, the authors find that particle softness correlates with
dynamic fragility, defined as the rate of increase of the structural relaxation time, ατ , with
particle concentration. Very soft particles have an Arrhenius-like (exponential in
concentration) growth of ατ , and increasing nonexponential growth (more fragile) as
particle stiffen. As a result, soft colloids appear as a promising system for understanding
the origin of fragility in glasses. Another long standing problem is the pursuit for a
unified picture for the ergodic-to-nonergodic glass transition in a wide range of soft
matter systems. Recent simulations and experiments [10] use microgels (or Hertzian
sphere modeling based on elastic contact mechanics [11]) as a pristine model system to
construct a jamming phase diagram [10]. A so called “soft jamming” transition has been
4
discovered corresponding to the non-monotonic change of the primary peak of the
interparticle pair correlation function, g(r), with increasing particle concentration.
Soft particle suspensions exhibit remarkable nonlinear shear rheology. They
respond either like an elastic solid when the applied stress is below the yield stress, or a
like a viscoelastic fluid when a stress greater than the yield value [12]. Thus they are very
important for processing and applications. It is well documented that dense suspensions
of soft particles are shear thinning fluids, and very often the shear stress increases with
the shear rate when above their yield stresses for the cases of concentrated emulsions [13-
15], microgel suspensions [16], and multilamellar vesicles [17]. Systematic rheological
studies of concentrated microgel suspensions, compressed emulsions [18-19], diblock
copolymer micellar solutions [20] and star polymers [21] have shown that the flow
properties of these materials are described by a universal flow curve.
All of the above novel phenomena call for a microscopic theory which can help us
to understand and provide materials design rules. Concerning microscopic theoretical
studies of soft repulsive sphere glass dynamics, there has been very little work. The only
study of glassy dynamics has applied standard ideal Mode Coupling Theory (MCT) [22],
which has several qualitative discrepancies: critical power law growth of relaxation times
with temperature and volume fraction instead of the observed strongly activated form
indicative of barrier hopping dynamics, and incorrect dynamic fragility trends with
particle stiffness. Such failures are no doubt due to the fact MCT does not capture
activated dynamics while the Nonlinear Langevin Equation (NLE) theory [23-26] is a
promising tool that includes the fundamentally important role of activated hopping events.
5
Adding nanoparticles or fillers to dense polymer melts can profoundly modify the
mechanical, thermal, optical and/or other material properties. These so-called polymer
nanocomposite (PNCs) are the subject of the second half of this thesis. They have a large
surface to volume ratio of nanoparticles which results in large modification of the matrix
polymer properties. Such property changes are strongly influenced by particle spatial
dispersion and PNCs structure, which is poorly understood. Recently, the use of nanogel
soft filler particles in PNCs revealed a host of intriguing phenomena, such as unexpected
high dispersion in the absence of polymer-particle attraction [27] (will be discussed in
Chapter 5), decrease of viscosity upon addition of the particles to the polymer matrix, and
striking particle size dependence of shear elasticity [28] (will be briefly discussed in
Chapter 7). These effects confirm the importance of these systems as rheological research
tools. To achieve such novel properties, the role of soft particle surface morphology must
be understood. The surface morphology can be viewed as a combination of surface
corrugation (roughness) and fluctuation (softness). Thus, a multi-scale modeling
approach is needed to fully understand the role of surface morphology at roughly the
monomer length scale which is most relevant to polymer mediated depletion attraction
forces between fillers which is a key driving force of undesirable macroscopic scale
clustering and phase separation .
The development and initial application of predictive theories of the equilibrium
structure and phase behavior of polymer nanocomposites based on hard filler has
undergone dramatic progress in the past decade [29]. However, much remains to be done,
and it seems fair to say that the development of broadly applicable microscopic theories
that can balance computation complexity inherent of real PNCs but still capture the key
6
physics is in its early stages. Many theoretical challenges and key questions remain open
for the PNCs based on the novel soft fillers described above, some of which are the
following. (1) What is the role of coarse graining of nanometer scale chemical structure ?
(2) How do different length scales interfere in dense PNCs? What are their roles in
nanoparticle dispersion? (3) How do the major states of polymer-mediated organization []
(depletion clustering, steric stabilization via polymer adsorption, and polymer-mediated
bridging and network formation) change for nonsmooth fillers of varying corrugated
surface, size and dynamic fluctuation? (4) How does the effective interfacial cohesion
between homopolymers and fillers change thermodynamics at different level of
description, i.e. from surface local interaction to particle center of mass level interaction?
(5) Does filler surface softness and shape fluctuations on the nanometer scale
perturbatively modify the states of organization and miscibility found for hard fillers, or
qualitatively new and unique behavior emerge?
The general goal of this thesis follows two tracks (see Fig.1.2). The first one
studies dense colloidal suspensions of soft particles. We first obtain the equilibrium
structure and then use it as an input to activated barrier hopping NLE theory to study the
slow activated dynamics, elasticity, and nonlinear rheology. The other track studies
polymer nanocomposites. First, we also obtain the equilibrium structure for two-
component system, and then we use it to study thermodynamics and equilibrium phase
diagrams.
7
1.2 Dissertation Overview
Chapter 2 presents the theoretical basis and methodology used to study the
equilibrium structure, glassy dynamics, and mechanical response of dense fluids of soft
repulsive particles, including Ornstein–Zernike integral equation theory to obtain
equilibrium liquid state structure, Naïve single particle MCT for the dynamical crossover,
and quiescent NLE theory for the activated hopping are summarized under quiescent and
stress-driven conditions. This methodology has been extensively developed in the group
and is general such that it can be applied to any spherical soft repulsive particle system.
Two representative types of model soft particle suspensions are studied in detail
in Chapters 3 (many arm stars) and 4 (Hertzian spheres mimicking crosslinked microgels)
applying the methodology of Chapter 2. We first consider the real space structural
consequences of single particle softness (interpenetrability for many arm stars and elastic
overlapping for Hertzian spheres) to study the “soft jamming” transition. Then we study
the NMCT dynamic crossover, glassy dynamical regime and activated barrier hopping,
and dynamic fragility. Lastly, calculations for the linear elasticity and nonlinear rheology
(for Hertzian sphere only) are presented. Qualitative and quantitative comparisons to
experiment are given.
In Chapter 5, a new hybrid small scale Monte Carlo (MC) simulation plus
Polymer Reference Interaction Site Model integral equation theory (PRISM) is proposed.
The methodology is applied to model soft particle polymer nanocomposites to understand
structure, thermodynamics and miscibility in Chapter 6. We first construct a minimalist
multi-scale model, which explicitly studies the role of surface corrugation and fluctuation
by introducing the dimensionless length scales for particle-monomer size asymmetry,
8
surface corrugation and fluctuation magnitude. Then we perform small-scale MC
simulation to correctly capture the effective interactions among particles and monomers.
We then employ PRISM theory to study the spatial statistical correlation functions of the
mixture to understand how the surface morphology affects depletion attraction and
bridging and whether it stabilizes dispersion.
Chapter 7 utilizes the multi-scale modeling developed in Chapter 5 in conjunction
with the Naïve MCT and NLE theory of dynamics and elasticity to study the shear
elasticity and glassy dynamics of dense fluids of soft nanoparticles of crosslinked
polymeric nature and understand the particle size dependence of kinetic vitrification and
the shear modulus.
Chapter 8 concludes the dissertation by summarizing the key results and briefly
sketches possible future directions.
1.3 References
[1] R.T. Bonnecaze and M. Cloitre, Adv Polym Sci (2010) 236: 117–161
[2] M.S.Wolfe, , “Dispersion and solution rheology control with swellable microgels,”
Prog. Org. Coat. 20, 487–500 (1992).
[3] Doi, M. and Edwards, S. F., The Theory of Polymer Dynamics, Oxford University
Press, New York (1986).
[4] de Gennes, P. G., Reptation of a Polymer Chain in the Presence of Fixed Obstacles, J.
Chem. Phys., 55 (1971) 572-579.
[5] Pusey, P. N., Colloidal Suspensions, In: Hansen, J. P., Levesque, D. and Zinn-Justin J.
(Eds.), Liquids, Freezing and Glass Transition, North Holland: Amsterdam (1991).
9
[6]. Cohen, E. G., de Schepper, I. M., Note on transport processes in dense colloidal
suspensions, J. Stat. Phys., 63 (1991) 241-248.
[7]. Frenkel J., Kinetic Theory of Liquids, Oxford, The Clarendon Press, (1946).
[8] Dimitris Vlassopoulos1, Emmanuel Stiakakis1 and Michael Kapnistos, Rheology
Reviews 2007 pp. 179-260
[9] J.Mattson, H.M.Wyss, A.Fernando-Nieves, K.Miyazaki, Z.Hu, D.R.Reichman and
D.A.Weitz, Nature, 462, 83 (2009).
[10] Z.Zhang, N.Xu, D.Chen, P.Yunker, A.M.Alsayed, K.B.Aptowicz, P.Habdas, A.J.Liu,
S.R.Nagel and A.G.Yodh, Nature, 459, 83 (2009).
[11] Liu, K. K., D. R. Williams, and B. J. Birscoe, “The large deformation of a single
microelastomeric sphere,” J.Phys. D 31, 294–303 _1998_.
[12] Barnes J (1999) J Non-Newton Fluid Mech 81:133
[13] Meeker SP, Bonnecaze RT, Cloitre M (2004) J Rheol 48:1295
[14] Goyon J, Colin A, Ovarlez G, Ajdari A, Bocquet L (2008) Nature 454:84
[15] Bécu L, Manneville S, Colin A (2006) Phys Rev Lett 96:138302
[16] Wyss H, Fernandez de Las Nieves A, Mattson J, Weitz DA (eds) (2010) Microgel
basedmaterials. Wiley-VCH, New York
[17] Fujii S, Richtering W (2006) Eur Phys J E 19:139
[18] Cloitre M, Borrega R, Monti F, Leibler L (2003) Phys Rev Lett 90:068303
[19] Seth JR (2008) On the rheology of dense pastes of soft particles. Ph.D. thesis, The
University of Texas
[20] Merlet-Lacroix N (2008) Auto-assemblage de copolymères à blocs acryliques en
solvent apolaire: un exemple de verre colloïdal attractif. Ph.D. thesis, University Paris VI
10
[21] Erwin BM, Cloitre M, Gauthier M, Vlassopoulos D (2010) Soft Matter doi:10.1039/
b926526k
[22] L.Berthier, E.Flenner, H.Jacquin and G.Szamel, Phys.Rev.E, 81, 031505 (2010).
[23] K.S.Schweizer and E.J.Saltzman, J.Chem.Phys., 119, 1181(2003).
[24] K.S.Schweizer, J.Chem.Phys., 123, 244501(2005).
[25] K.S.Schweizer, Curr.Opin.Coll.Inter.Sci., 12, 297(2007).
[26] E.J.Saltzman, K.S.Schweizer,Phys.Rev.E 74,061501 (2006);77, 051504 (2008).
[27] M.E.Mackay, A.Tuteja, P.M.Duxbury, C.J.Hawker, B. Van Horn and Z.H.Guan,
Science, 311, 1740 (2006).
[28] A. TUTEJA, M.E. MACKAY, C. J. HAWKER, B. VAN HORN, D. L. HO, Journal
of Polymer Science: Part B: Polymer Physics
[29] Lisa M. Hall b, Arthi Jayaraman c, Kenneth S. Schweizer, Current Opinion in Solid
State and Materials Science 14 (2010) 38–48
11
1.4 Figures
Fig. 1. 1
Schematic representation of the soft particle family. The two types of soft particles in red
box are of major concerns for this thesis. Original figure from Bonnecaze and Cloitre,
Adv. Polym. Sci., 236 (2010).
12
Fig. 1. 2
Schematic representation of the general goal of thesis: for colloidal suspension (top), we
first obtain the equilibrium structure and then use it as a input to activated barrier hopping
and mechanical response theories. For polymer nanocomposites, we also first determine
its equilibrium structure then use it to study thermodynamics and phase diagram.
13
CHAPTER 2
GLASSY DYNAMICS AND MECHANICAL RESPONSE IN
DENSE FLUIDS OF SOFT REPULSIVE SPHERES:
METHODOLOGY
2.1 Introduction
The hard sphere fluid or colloidal suspension is a pristine model system to
investigate glassy dynamics due solely to singular repulsive excluded volume forces and
crowding [1,2]. Many ensemble-averaged aspects are well described by ideal mode
coupling theory (MCT) [2] in the initial (or precursor) slowing down regime that covers a
few orders of magnitude of relaxation based on adjustment of the critical nonergodicity
volume fraction. Ideal MCT self-consistently describes collective time-dependent density
fluctuations based on the confining cage concept and continuous small amplitude
cooperative motions within a dynamical Gaussian approximation [2]. However, confocal
microscopy [3] and computer simulations [4] both find single particle trajectories display
intermittent large amplitude activated hopping events at high volume fractions, even in
the dynamical precursor regime. A recent experimental and simulation study [5] suggests
activated dynamics is dominant well below random close packing (RCP). The many
highly nongaussian phenomena observed [6] are not captured by ideal MCT, but are well
described by the nonlinear Langevin equation theory (NLE) as a consequence of
thermally activated barrier hopping and intermittent trajectories [6-9]. Hence, the
prediction of ideal MCT, a dynamical Gaussian approach that ignore large amplitude
hopping of complete arrest is now understood to actually indicate a dynamic crossover
14
from collective but smooth Gaussian like dynamics to a slower form of activated
transport and relaxation.
Perhaps the simplest variation away from hard spheres are soft repulsive spherical
particles. Experimentally, distinct families of such colloids or nanoparticles have been
created where the degree of softness, and functional form of the interparticle potential,
are different and widely tunable. For example, recent experimental studies have appeared
for dense suspensions of many arm star polymers [10,11], crosslinked microgels [12-19],
and copolymer micelles [20]. The repulsive interaction between many arm stars in a good
solvent are mediated by the underlying polymer chains, and at the center-of-mass level
consist of an arm-number-dependent longer range Yukawa repulsion and a logarithmic
repulsion at small separations (see Fig.2.1). In contrast, a popular simple model for
microgels is the finite range, repulsive Herztian contact interaction [21,22] which remains
finite at full particle overlap (see Fig.2.1). This potential, or minor variants of it, have
also seen much simulation study as a model system for athermal granular jamming [23].
What all dense fluids of soft repulsive particles have in common is the emergence of
intermittent dynamics and glass-like viscoelasticity at high enough concentrations with
qualitatively distinct features not present in hard sphere suspensions. Besides the broad
relevance to materials science, soft particles are ubiquitous in other areas such as polymer
nanocomposites [24] and vesicle diffusion in biopolymer networks [25].
Recent experiments have uncovered a wealth of fascinating phenomena including:
(i) nonArrhenius variation of the relaxation time with particle stiffness (or effective
inverse temperature) and fluid volume fraction [13,14], (ii) massive variability of the
dynamic fragility associated with relaxation and viscous flow [13,17], (iii) non-
15
monotonic variation of local packing structure with volume fraction due to the avoidance
of strict jamming (so-called “soft jamming crossover”) [14], (iv) a direct connection
between vitrification volume fraction and the elastic shear modulus [12], (v) particle
softness-independent linear scaling of the shear modulus with concentration at ultra-high
(beyond the hard sphere RCP) volume fractions [17,21], and (vi) distinctive dependences
of the yield stress and other nonlinear rheological properties on volume fraction and
degree of particle softness [26].
Recent computer simulation studies of the quiescent dynamics and elasticity of
the soft repulsion models have observed some of the above features [17,21,27]. An
interesting two-branch universality of the temperature and concentration dependences of
the mean alpha relaxation time has been uncovered [27]. More generally, recent
theoretical and simulation studies have also discovered a rich equilibrium phase behavior,
including re-entrant phase transitions at ultra-high volume fractions due to subtle
competition of energetic and entropic factors associated with soft particle overlap [22].
However, it is important to emphasize that the simple center-of-mass central force
model potentials employed in simulation and theoretical studies [22] neglect a host of
complexities present in polymer-based microgels (and also stars, micelles, grafted
nanoparticles, emulsion droplets, liposomes [17]) which are especially relevant at ultra-
high concentrations. For example, complex monomer chemistry, dangling (perhaps
entangled) chains at the particle surface, residual attractions, and Coulomb repulsions
(polyelectrolyte effects) can be present, along with distinctive consequences of shape
deformability above random close packing such osmotic deswelling and the formation of
solvent-lubricated facets [17]. These complications also render a unique experimental
16
quantification of volume fraction difficult. Hence, simulation and theoretical studies
based on the simple interaction potentials are not generally expected to quantitatively
describe suspensions of soft particles, but rather serve as a zeroth order description of
many (not all) aspects at a semi-quantitative or qualitative level.
Concerning microscopic theoretical studies of soft repulsive sphere glass
dynamics, there has been very little work. A micromechanical model has been developed
for the elastic and flow properties beyond contact (volume fractions RCP
φ ~0.64) [17, 21].
The only study of glassy dynamics has applied standard ideal MCT [28]. Ideal MCT does
predict a form of dynamic scaling (over a tiny volume fraction range) based on the
hypothetical ideal glass transition singularity, and a critical power law parabolic form of
the effective dynamic crossover temperature, as seen in the simulations. However, several
qualitative discrepancies were clearly evident: critical power law grow of relaxation
times with temperature and volume fraction instead of the observed strongly activated
form, and incorrect dynamic fragility trends with particle stiffness. Such failures are no
doubt due to the fact MCT does not capture activated dynamics.
Hence, the intriguing experimental and simulation glassy dynamics observations
are not theoretically understood in a consistent manner. We believe this reflects the
fundamentally important role of activated hopping events. This motivates our present
work which comprehensively applies the existing NLE theory of Schweizer and
coworkers, which at the moment seems to be the only first principles no adjustable
parameter theory of activated glassy dynamics and viscoelasticity at the level of forces, to
the models of soft repulsive sphere fluids. Relaxation, elasticity, and some aspects of
nonlinear rheology are addressed within a single framework. Given the interaction
17
potential uncertainties for real soft particles, we do not emphasize quantitative
comparisons to experiment, but rather only qualitative comparisons.
Our goal is to apply and generalize existing statistical mechanical methods to
dense fluids of soft particles to understand how particle softness affects structure and
dynamics. But in order to create a predictive theory, one needs to relate structure, forces,
and dynamics. We first use the model interparticle potentials between soft particles,
which are controlled by chemistry, as an input to well-known equilibrium liquid state
theory and predict equilibrium structure, which can be quantified via the pair correlation
function and static structure factor as an input to our dynamical theories to address the
following issues: (1) dynamic free energy and dynamic crossover, (2) the influence of
volume fraction and particle softness on the mean relaxation time, (3) kinetic arrest
diagram and dynamic fragility, (4) the quiescent shear modulus as a function of volume
fraction and repulsion strength and (for microgel models) (5) nonlinear rheology. The
focus of chapter 3 is on many arm stars, while chapter 4 studies crosslinked microgels.
The remainder of this chapter is structured as follows. Section 2.2 discusses the
equilibrium liquid sate Ornstein-Zernike theory. The single particle “naïve” MCT
dynamical crossover is given in section 2.3. The quiescent NLE approach is reviewed in
section 2.4. The ultralocal limit is introduced in Section 2.5. Section 2.6 briefly reviews
the NLE and naïve MCT approaches employed to determine mechanical properties.
Section 2.7 and 2.8 present the references and figures.
2.2 Equilibrium Liquid State Theory: Ornstein–Zernike Theory
18
Our interest is to use the interparticle pair potential as input to a theory for the
spatial structure at the center of mass level of the soft particle fluid at nonzero
concentrations. We employ the Ornstein–Zernike (OZ) approach [28] based on a single
integral equation that relates the interparticle pair correlation function, ( ) ( ) 1h r g r≡ − , to
the direct correlation function, ( )C r , as
' ' '( ) ( ) (| |) ( )h r C r C r r h r drρ= + −∫
(2.1)
where r is the CM distance between two particles and ρ is the particle number density.
The Fourier space static structure factor that quantifies collective density fluctuations is
given by:
1
( ) 1 ( )1 ( )
S k h kC k
ρρ
= + =−
(2.2)
To solve the OZ equation a closure approximation is required. For separations
beyond the distance of closest approach ( σ , the particle diameter), we employ the
hypernetted chain (HNC) closure for dense fluids of soft particles:
( ) ( ) ln( ( )) ( )C r U r g r h rβ= − − + (2.3)
where U(r) is the interparticle potential. For r D< , the hard core exclusion constraint is
exactly enforced, ( ) 0g r = . We adopt the HNC closure for multiple reasons: (1) it
ensures positivity of g(r) for all separations, (2) it guarantees the correct result is obtained
in the dilute limit, and (3) the HNC has proved accurate and reliable for many soft
repulsive systems at the CM level (e.g. finite range Hertzian repulsive spheres [29] and
carbon black fractals [30]). The OZ-HNC integral equation is solved using the KINSOL
algorithm (hybrid Picard plus Newton–Raphson method) [31].
19
2.3 Naïve MCT Dynamical Crossover
We determine the ideal MCT glass transition, which represents a dynamic
crossover to activated dynamics, using the single particle “naïve” MCT (NMCT) [6,32].
This mathematically and physically simple approach is based on computing the long time
limit of the force-force correlation function, ( )K t , given in a Fourier-resolved form by:
( )2 2 11 ( ) / 62 2
2 3
1( ) (0). ( ) ( ) ( )
3 (2 )
lock r S kdk
K t F F t k C k S k eρβ π
−− +→ ∞ = • → ∞ = ∫
(2.4)
where ( )F t
is the total force exerted on a tagged particle by the surrounding fluid at
time t , rloc is the long time limit of a particle mean square displacement or “localization
length”, and a Gaussian Einstein solid description of the arrested glass is adopted. If rloc is
infinite, the system is a fluid; if it is non-zero, an ideal glass. The dynamical “vertex”,
defined as 2 2 ( ) ( )k C k S kρ , describes the Fourier resolved effective mean square force on
a particle. The localization length obeys a nonlinear self-consistent equation [6]:
2 2 1[1 ( )]/ 62 2
2 3
1 1( ) ( )
9 (2 )loc
k r S k
loc
dkk C k S k e
rρ
π
−− += ∫
(2.5)
We recall that NMCT has successfully captured many subtle dynamical arrest
phenomena including: re-entrant glass melting and subsequent attractive glass formation
in sticky hard spheres [33], non-monotonic variation of glass volume fraction with aspect
ratio in fluids of diatomic and other uniaxial hard objects [34], and re-entrant phenomena
and single versus double localization behavior in mixtures of repulsive and attractive
particles [35].
2.4 Quiescent NLE Theory
20
To go beyond ideal MCT and treat single particle activated dynamics we employ
the microscopic force-level NLE theory [6,7]. This approach is built on a locally solid-
state view of highly viscous dynamics with density fluctuations the key slow variable,
and a local equilibrium approximation to mathematically close the theory at the single
particle motion level. NLE theory is formulated as a stochastic equation-of-motion for the
instantaneous scalar displacement of a tagged particle, r(t):
( ( ))( )( )
( )
eff
S
F r tr tf t
t r tζ δ
∂∂= − +
∂ ∂ (2.6)
( )2 2
121 ( )
63 1
( ) ( )( ) 3ln( )
(2 ) 1 ( )
k rS k
eff
dk C k S kF r r e
S k
ρβ
π
−− +
−= − −
+∫
(2.7)
Here, /S B S
k T Dζ = is a short time friction constant, the random thermal force
satisfies (0) ( ) 2 ( )B S
f f t k T tδ δ ζ δ< >= , and ( ( ))eff
F r t is a “dynamic free energy” the
derivative of which quantifies the force on a tagged particle due to its surroundings. As
confining forces increase, Feff(r) changes from a delocalized form (decreasing function of
r), to a transient localized form (local minimum at rL with a barrier of height FB)
indicating a crossover to activated dynamics. If the thermal noise in Eq. (2.6) is dropped
(no hopping), or a literal dynamical Gaussian approximation is made at the ensemble-
averaged level, the NMCT transition [7] is obtained which can be mathematically
expressed as the first emergence of a localized solution of ( )eff / | 0L
r rF r r =∂ ∂ = .
Activated hopping or relaxation is quantified via the Kramers mean first passage
time for barrier crossing:
( )1/ 2
0/ 2 exp( )hop s B B
K K Fτ τ π β−
= (2.8)
21
where 2
s sτ βσ ζ= is the elementary short time scale, and 0K and
BK the well and barrier
curvatures in units of 2
Bk Tσ − , respectively [6]. The hopping time is almost identical to
the relaxation time of the experimentally measurable incoherent dynamic structure factor
at the cage peak of S(k) [9], a common measure of “structural relaxation”.
We mention that the single particle NLE theory has very recently been
generalized to treat the space-time correlated dynamics of a pair of particles [36].
Applications of this more sophisticated NLE theory to study average relaxation questions
for hard sphere fluids reveals good agreement with the simpler single particle approach,
thereby providing deeper theoretical support for the practical usefulness of the latter.
2.5 Ultralocal Limit
The structural input to NMCT and NLE theory that quantifies the Fourier-
resolved mean square force on a tagged spherical particle is given by the “vertex” [37]:
4 2( ) ( ) ( )V k k C k S kρ= (2.9)
When barriers are high, it is the analytically known that for interaction potentials with a
hard core (hence a rigorous exclusion constraint on g(r)) that the large wavevector limit
of Eq(6) dominates all predictions of NLE theory, from the localization length to the
barrier height [38,39]. Specifically, the vertex becomes k-independent at large
wavevectors, with an amplitude proportional to a “coupling constant”, λ . Physically, the
latter is a measure of the dynamical mean square force exerted on a tagged particle due to
repulsive force collisions, and for spheres is:
2 ( )gφλ = σ (2.10)
22
where ( )g σ is the contact value. Prior work for hard core models [34,38,39] has found
the barrier height is proportional to this coupling constant once B
F exceeds a few kBT:
( ) /B c c
F ∝ λ − λ λ (2.11)
where c
λ is the coupling constant at the ideal NMCT transition. Moreover, it has very
recently been shown that for broad families of nonspherical particles with hard core
repulsions, with and without attractions, Eq. (2.11) is also remarkably accurate [40].
For the soft repulsive sphere systems of present interest, which do not have a hard
core, the prior analytic analysis does not rigorously apply. Moreover, exact analytic exact
knowledge of the relevant direct correlation function of these systems in the high k limit
is not available. But, on general grounds the vertex in Eq. (2.9) is expected to behave
differently at large wavevectors since there are no discontinuous jumps of g(r) to zero as
for systems with a hard core repulsion. We shall empirically test in the following two
chapters whether the basic ideas of the hard core based analytic limit of NLE apply to
soft particles. But in stead of using the contact value ( )g σ , we use 1g , the primary peak
value instead. And thus Eq.(2.10) becomes 2
1gφλ = .
2.6 Mechanical Response
The linear elastic shear modulus, G’, is computed using the standard statistical
mechanical formula where stress is projected onto bilinear collective density fluctuation
modes and 4-point correlation functions are factorized [41-43]:
2 2
2
/3 ( )2
2
0
' ln( ( ))60
LOCk r S kB
k T dG dk k S k e
dkπ
∞−
= ∫ (2.12)
23
Here, S(k) is the fluid static structure factor, and rLOC is the minimum of the dynamic free
energy (transient localization length) computed from the naïve MCT self-consistency
Eq.(2.6). Eq.(2.12) literally applies only in the ideal nonergodic state of MCT. We
employ it here as a sensible approximation of the physically relevant intermediate time
plateau of the dynamic stress relaxation function, G(t), associated with the transiently
localized state which exists even in the presence of activated hopping at long times.
For hard sphere fluids [44], and also fluids with an attraction plus a hard core
repulsion [45], an analytic relation between the glassy shear modulus and the localization
length has been derived:
2
18'
5
B
LOC
k TG
r
φ
σπ (2.13)
where σ is the particle diameter. As will be discussed in section 2.3, we find Eq.(2.13)
describes very well our numerical calculations for the repulsive Hertzian fluids.
Moreover, Eq.(2.13) has been recently shown experimentally to be remarkably accurate
for both polymer-colloid depletion gels [46] and microgel suspensions [47,48].
The extension of NLE theory to nonlinear response has been achieved for
spherical particle systems [49], and extensively applied to hard [49-51] and attractive
[52,53] spheres, based on an applied stress and single particle microrheology perspective.
Partial motivation for the theoretical approach comes from simulation findings that local
dynamics is accelerated in a nearly isotropic manner despite the anisotropic nature of the
macroscopic deformation, and local single particle (or segmental for polymers) dynamics
is tightly coupled with bulk mechanical response [54-58]. Indeed, simulations find
massive shear thinning can occur with little or no anisotropy of local structure, i.e. the
system remains “effectively isotropic” on the cage scale [55,56,58]. The recent
24
theoretical/simulation studies of Brady and coworkers [59] also supports the assumption
that bulk mechanical response can be accurately determined from single colloid dynamics
(microrheology) even when the tagged and matrix particles are identical in size.
The specific ansatz underlying the nonlinear NLE approach is that a macroscopic
applied stress (magnitude τ ) induces a constant scalar external force on a tagged
segment given by: 2f cσ τ= , where c is an of order unity and weakly volume fraction
dependent factor estimated from an effective particle cross-sectional area [49]. Applied
deformation then modifies the dynamic free energy as [42,49,52]:
2 2/3
( ; ) ( ; 0)eff eff
F r F r fr
f a
τ τ
τσ φ −
= = −
= (2.14)
More explicitly, using Eq.(2.7), one has
( )2 2
121 ( )
2/3 263 1
( ) ( )( ) 3ln( )
(2 ) 1 ( )
k rS k
eff
dk C k S kF r r e a r
S k
ρβ βφ σ τ
π
−− +−
−= − − −
+∫
(2.15)
where here we employ a= / 6π [49]. External stress enters ala an instantaneous
dynamical variable analog of the Eyring [60] landscape tilting idea. The physical picture
is stress introduces a bias which lowers activation barriers and increases localization
length corresponding to a direct mechanical acceleration of relaxation and reduction of
rigidity. The stress-induced force is assumed to not modify equilibrium pair correlations
(which are input to the dynamic free energy) nor the fast (very local) relaxation process
sτ in the Kramers time (Eq.(2.8)). Volume-changing versus volume-conserving
deformations are not distinguished in this simple approach.
Stress lowers the barrier in the dynamic free energy, and can mechanically drive a
glass-to-liquid transition in a manner consistent with potential energy landscape
25
simulations [61]. However, the consequences of this stress-assisted speeding up of
relaxation within NLE theory display multiple non-Eyring features [42,49], as observed
in many simulations and experiments [54,62-64]. The localization well and barrier of the
dynamic free energy are destroyed at an “absolute yield stress”, abs
τ . Here, cage escape
no longer requires thermal activation ala an “athermal” or “granular” limit. Stress also
increases the localization length, thereby reducing the elastic modulus in Eq.(2.12).
A simple constitutive equation has been constructed in the Maxwell model spirit
which requires as input only the elastic modulus and mean segmental relaxation time
[42,65]. Here we focus solely on the two extreme limits which do not require the full
constitutive equation. The first is a step strain deformation under conditions where there
are no thermally-induced activated hopping on the experimental time scale. This
corresponds to an elastic solid-like rheological equation-of-state [49]:
'( )Gτ τ γ= (2.16)
Eq.(2.16) implicitly defines strain, γ , and the nonlinear shear modulus follows from
Eq.(2.12) where stress enters only via an increase of the localization length. Since abs
τ is
the minimum stress required to destroy the activation barrier, it is equivalent to the
condition that the applied force, f, is equal and opposite to the maximum cage restoring
force of the quiescent system computed from the dynamic free energy. The “absolute
yield strain” is:
, / '( )y abs abs abs
Gγ τ τ= (2.17)
A “mixed” yield strain, employed in some experimental studies, is also defined:
, / '(0)y mixed abs
Gγ τ= (2.18)
26
The second limit of interest is the long time flow regime. Based on a Maxwell
model, which is useful when barriers are high and the short time (bare) contribution to the
viscosity (η ) is not important, one has the steady state flow relation:
( ) '( ) ( )Gτ γ η τ γ τ τ τα= = (2.19)
where the (shear) deformation rate is γ , and the stress-dependent mechanical relaxation
time is identified as the mean hopping time of Kramers theory [49]:
( )
0
( ) 2
( ) ( )
BF
sB
e
K K
τατ τ π
τ τ τ=
(2.20)
Applied stress distorts all aspects of the dynamic free energy. The shear rate, γ , follows
from the viscous flow relation / ( )γ τ η τ= , thereby allowing calculation of the shear
thinning response , i.e., ( ), ( )ατ γ η γ . We note that prior predictions of ( )ατ γ for a single
tagged colloid in a glassy hard sphere fluid under strong shear [49] have been
quantitatively tested against suspension experiments [66]. Excellent, nearly quantitative
agreement with experiment was demonstrated [42]. In many other non-glassy complex
fluids, perhaps most notably concentrated polymer solutions or melts not in the
supercooled regime [67], shear thinning emerges when the deformation and quiescent
relaxation time scales are comparable, i.e. the Peclet number is of order unity,
(0) 1Pe αγτ= ≈ . In contrast, the present theory predicts shear thinning begins in glassy
particle fluids at extremely small values of the Peclet number, (0) 1Pe αγτ= << . The
physical picture is shear thinning occurs not via destruction of the barrier, but rather by a
stress-induced reduction of the barrier and corresponding massive increase of the
activated hopping rate that permits flow on the experimental time scale (inverse shear
27
rate). It is this mechanism that leads to the onset of shear thinning at remarkably small
Peclet numbers, as observed recently in computer simulation [68]. The physical picture of
flow via stress-accelerated activated events is also consistent with the experimental
confocal data for particle trajectories [66].
2.7. References
[1] W.M.van Megen and S.Underwood, Phys.Rev.E, 47, 249 (1993).
[2] W.Gotze,J.Phys.:Condens.Matter,11,A1(1999).
[3] E.R.Weeks and D.A.Weitz, Chem.Phys., 284, 361 (2002).
[4] S.K.Kumar, G.Szamel, and J.F.Douglas, J.Chem.Phys.,124,214501(2006);
Y.Brumer and D.R.Reichman, Phys.Rev.E, 69, 041202 (2004).
[5] G.Brambilla, D.ElMasri, M.Pierno, L.Berthier, L.Cipelletti, G.Petekidis and
A.B.Schofield, Phys.Rev.Lett., 102, 085703 (2009).
[6] K.S.Schweizer and E.J.Saltzman, J.Chem.Phys., 119, 1181(2003).
[7] K.S.Schweizer, J.Chem.Phys., 123, 244501(2005).
[8] K.S.Schweizer, Curr.Opin.Coll.Inter.Sci., 12, 297(2007).
[9] E.J.Saltzman, K.S.Schweizer,Phys.Rev.E 74,061501 (2006);77, 051504 (2008).
[10] F.Ozon, G.Petekidis and D.Vlassopoulos, Ind. Eng. Chem. Res.,45, 6946 (2006);
M.E.Helgeson, N.J.Wagner and D.Vlassopoulos, J.Rheology, 51, 297 (2007).
[11] B.M.Erwin, M.Cloitre, M.Gauthier and D. Vlassopoulos, Soft Matter, 12, 2825
(2010); B.M.Erwin, D.Vlassopoulos and M.Cloitre, J.Rheology, 54, 915 (2010).
[12] E.H.Purnomo, D.van den Ende, S.A.Vanapili and F.Mugele, Phys.Rev.Lett. 101,
23801 (2008).
28
[13] J.Mattson, H.M.Wyss, A.Fernando-Nieves, K.Miyazaki, Z.Hu, D.R.Reichman and
D.A.Weitz, Nature, 462, 83 (2009).
[14] Z.Zhang, N.Xu, D.Chen, P.Yunker, A.M.Alsayed, K.B.Aptowicz, P.Habdas, A.J.Liu,
S.R.Nagel and A.G.Yodh, Nature, 459, 83 (2009).
[15] H.Senff and W.Richtering, J.Chem.Phys., 111, 1705 (1999).
[16] M.Cloitre, R.Borrega, F.Monti and L.Leibler, Phys.Rev.Lett., 90,068303 (2003).
[17] For excellent recent reviews of microgel suspensions, see: M.Cloitre, in “Microgel-
based Materials”, H.Wyss, A.Fernandez de las Nieves, J.Mattson and D.A.Weitz Eds;
Wiley-VCH, 2010; R.Bonnecaze and M.Cloitre, Adv.Polym.Sci, 236, 117 (2010).
[18] B.R.Saunders and B.Vincent, Adv.Colloid Interface Sci. , 80, 1, (1999)
[19] M.S.Wolfe and C.Scopazzi, J.Colloid and Interface Sci., 122, 265 (1989).
[20] W.R.Chen, F.Mallamace, C.J.Glinka, E.Fratini and S.H.Chen, Phys.Rev.E,
68,041402 (2003).
[21]J.R.Seth, M.Cloitre and R.T.Bonnecaze,J.Rheology, 50, 353 (2006); 52, 1241 (2008).
[22] C.Likos, Soft Matter, 2, 478 (2006); C.Likos, Phys.Rep., 348, 267 (2001).
[23] C.O’Hern, L.E.Silbert, A.J.Liu and S.R.Nagel, Phys.Rev.E, 68, 011306 (2003).
[24] M.E.Mackay, A.Tuteja, P.M.Duxbury, C.J.Hawker, B. Van Horn and Z.H.Guan,
Science, 311, 1740 (2006).
[25] C. P. Brangwynne, G. H. Koenderink, E. Barry, Z. Dogic, F. C. MacKintosh and D.
A. Weitz, Biophys. J. 93, 346 (2007).
[26] A.LeGrand and G.Petekidis, Rheol.Acta, 47, 579 (2007); V.Carrier and G.Petekidis,
J.Rheology, 53, 245 (2009).
29
[27] L.Berthier and T.A.Witten, Europhys.Lett., 86, 10001 (2009); Phys.Rev.E, 80,
021502 (2009).
[28] L.Berthier, E.Flenner, H.Jacquin and G.Szamel, Phys.Rev.E, 81, 031505 (2010).
[29] H.Jacquin and L.Berthier, Soft Matter, 6, 2970 (2010)
[30] R.Chakrabarti, Jean-Yves Delannoy, M. Couty and K.S. Schweizer, Soft Matter, 7,
5397 (2011)
[31] L. M. Hall and K. S. Schweizer, J. Chem. Phys., 2008, 128, 234901
[32] T.R.Kirkpatrick and P.G.Wolynes, Phys. Rev.A, 35, 3072 (1987).
[33] Y.-L. Chen and K.S.Schweizer, J.Chem. Phys., 120, 7212 (2004).
[34] G.Yatsenko and K.S.Schweizer, Phys. Rev. E, 76, 041506 (2007); R.Zhang and
K.S.Schweizer, Phys. Rev. E, 80, 011502 (2009).
[35] D.Viehman and K.S.Schweizer, J.Chem. Phys., 128, 084508 (2008).
[36] D.M.Sussman and K.S.Schweizer, J.Chem.Phys., 134, 064516 (2011).
[37] K.S.Schweizer and G.Yatsenko, J.Chem. Phys., 127, 164505 (2007).
[38] K.S.Schweizer and G.Yatsenko, J.Chem. Phys., 127, 164505 (2007).
[39] K.S.Schweizer, J.Chem.Phys.,127, 16506 (2007).
[40] M.Tripathy and K.S.Schweizer, Phys.Rev. E, 83, 041406 and 041407 (2011).
[41] W.Gotze, J.Phys.:Condens.Matter, 11, A1(1999).
[42] K.Chen, E.J.Saltzman and K.S.Schweizer, J. Phys.-Condensed Matter, 21, 503101
(2009); Annual Reviews of Condensed Matter Physics, 1, 277 (2010).
[43] Y.-L. Chen and K.S.Schweizer, J.Chem. Phys., 120, 7212 (2004).
[44] K.S.Schweizer, J.Chem.Phys., 123, 244501(2005).
[45] M.Tripathy and K.S.Schweizer, Phys.Rev. E, 83, 041406 and 041407 (2011).
30
[46] H.Y.Guo, S.Ramakrishnan, J.L.Harden and R.L.Leheny, Phys.Rev.E, 81,050401
(2010).
[47] M.Cloitre, Macromol.Symp., 229, 99 (2005).
[48] For excellent recent reviews of microgel suspensions, see: M.Cloitre, in “Microgel-
based Materials”, H.Wyss, A.Fernandez de las Nieves, J.Mattson and D.A.Weitz Eds;
Wiley-VCH, 2010; R.Bonnecaze and M.Cloitre, Adv.Polym.Sci, 236, 117 (2010).
[49] V.Kobelev and K.S.Schweizer, Phys. Rev. E, 71, 021401 (2005).
[50] R.Rao, V.Kobelev, Q.Li, J.A.Lewis and K.S.Schweizer, Langmuir, 22, 2441 (2006).
[51] E.J.Saltzman, G.Yatsenko and K.S.Schweizer, J.Phys.-Condensed Matter, 20,
244129 (2008).
[52] V.Kobelev and K.S.Schweizer, J.Chem. Phys., 123, 164902 and 164903 (2005).
[53] V.Gopalakrishnan, K.S.Schweizer and C.F.Zukoski, J.Phys.-Condensed Matter, 18,
11531 (2006).
[54] M.Warren and J.Rottler, Phys.Rev.E, 76, 031802 (2007).
[55] R.Yamamoto and A. Onuki, Phys.Rev.E, 58, 3515 (1998).
[56] K.Miyazaki, R.Yamamoto and D.R.Reichman, Phys.Rev.E, 70, 011501 (2004).
[57] R.A.Riggleman, H.N.Lee, M.D.Ediger, and J.J.dePablo, Phys.Rev.Lett., 99 215501
(2007).
[58] R.A.Riggleman, K.S.Schweizer, and J.J.dePablo, Macromolecules, 49 4969 (2008) .
[59] A.S.Khair and J.F.Brady, J.Fluid Mechanics, 557, 73 (2006); A.S.Khair and
J.F.Brady, J.Rheology, 49, 1449 (2005); I.C.Carpen and J.F.Brady, J.Rheology, 49, 1483
(2005).
[60] H.Eyring, J.Chem.Phys. 4, 283 (1936).
31
[61] M.J.Osbourne and D.J.Lacks, J.Phys.Chem.B 108, 199619 (2004); D.L.Malandro
[62] J.Rottler and M.O.Robbins, Phys.Rev.Lett., 95, 225504 (2005).
[63] J. Rottler and M.O.Robbins, Phys.Rev.E, 68, 011507 (2003).
[64] B,Vorselaars, A.V.Lyulin and M.A.J.Michels, J.Chem.Phys., 130, 074905 (2009).
[65] K.Chen and K.S.Schweizer, Macromolecules, 41, 5908 (2008); K.Chen,
K.S.Schweizer, R.Stamm, E.Lee and J.M.Caruthers, J.Chem. Phys., 129, 184904 (2008).
[66] R.Besseling, E.R.Weeks, A.B.Schofeld and W.C.K.Poon, Phys.Rev.Lett. 99, 028301
(2007).
[67] See, for example, M.Doi and S.F.Edwards, The Theory of Polymer Dynamics
(Oxford University Press, London, 1986).
[68] A. Furakawa and H.Tanaka, private communication, 2010.
32
2.8 Figures
Fig. 2. 1
Above: Schematic representations of many arm star polymers and crosslinked microgels
nanoparticles are shown. Below: The corresponding center of mass level interparticle
potentials.
33
CHAPTER 3
GLASSY DYNAMICS AND ELASTICITY OF
DENSE FLUIDS OF MANY ARM STARS
3.1 Introduction
Many arm star polymers with tunable number and size of arms, and thus
interactions, represent ideal model systems for exploring the regime of soft material
behaviour that interpolates between hard spheres and polymeric coils. They can be
thought of as ultrasoft colloidal spheres with a very small deformable core and a corona
consisting of grafted chains (arms) (see Fig.2.1). This regime is characterized by a rich
variety of properties that reflect a combination of polymeric and colloidal features.
In this chapter, we apply the methodology in chapter 2 to perform the first
theoretical study of the four colloidal characteristic features of soft particles discussed in
the Introduction of Chapter 2 in the context of dense solutions of many arm stars. These
nanocolloids are also a model system for other soft particles such as polymer-grafted
colloids. Dramatic non-hard-sphere dynamical effects are predicted, which are
understandable based on a connection between local packing structure and caging forces.
Section 3.2 discusses the model, equilibrium structure, and “soft jamming” crossover. In
section 3.3, we study the Naïve MCT kinetic arrest map. Section 3.4 focuses on activated
barrier hopping and dynamical fragility. In section 3.5, we study the linear elastic shear
modulus and experiment-theory confrontations are shown. Section 3.6 uses the ultralocal
limit analysis to understand the behaviors of dynamic barrier height, mean barrier
hopping time and linear elastic modulus. The chapter concludes in section 3.7 with a
34
summary and discussion. Section 3.8 and 3.9 present the references and figures,
respectively.
3.2 Model, Equilibrium Structure and “Soft Jamming” Crossover
We adopt the well studied center-of-mass (CM) description of stars composed of f
polymer arms in an athermal good solvent that interact via a pair decomposable repulsion
of logarithmic (Yukawa) form at small (large) separations:
( )
( ) ( )
13/ 2
13/ 2
18 ( ) / 5 [ ln( / ) 1 / 2 ] ,
/ 1 / 2 exp[ ( ) / 2 ] ,
V r f r f r
f r f f r r
β σ σ
σ σ σ σ
−
−
= − + + ≤
= + − − >
(3.1)
where r is the CM separation, σ ~1.28Rg (Rg is radius of gyration), and 1
Bk T β −=
is
thermal energy [1-3]. The strength and spatial range of V(r) depends strongly on arm
number, and hard spheres are recovered as f → ∞ . Star concentration is quantified via a
volume fraction 3 / 6φ ρπσ= , where ρ is the particle number density. Comparison of
scattering experiments with the predictions of simulation and integral equation theory for
the structural correlations of dense many arm star polymer fluids based on Eq.(3.1) show
good agreement. Moreover, monomer level simulations of the inter-star CM potential-of-
mean-force agree well with Eq.(3.1) up to high reduced concentrations ( 1φ > )[1]. We
employ the Ornstein-Zernike equation with HNC closure [4] to compute the structure
factor ( )1
( ) 1 ( ) 1 ( )S q C q h qρ ρ−
= − = + , where C(q) and h(q) are the Fourier transform of
the direct correlation function and nonrandom pair correlation function, h(r)=g(r)-1,
respectively [22].
35
We first consider the real space structural consequences of star interpenetrability.
Fig.3.1 shows representative results for the pair correlation function of a f=128 arms star.
With increasing volume fraction, the cage peak on the star size scale (r~σ ) initially
increases and shifts to smaller separations. However, beyond 0.55J
φ φ≅ ≡ (weakly f-
dependent), qualitative changes emerge: the primary ordering peak decreases with
concentration and a new peak appears at a much smaller (overlapped) CM separation,
effects due to star interpenetration and increasing dominance of the logarithmic repulsion.
We refer to J
φ as a “soft jamming” threshold in analogy with a recent study of microgel
suspensions [8]. This interpretation follows from the inset which shows the cage peak of
g(r) goes through a maximum atJ
φ . This non-monotonic behavior has been observed in
experiment [8], simulation [8], and HNC integral equation theory studies [7] for repulsive
microgels, and will be shown below to have dramatic consequences on the elastic and
dynamic properties of dense many arm star fluids. We note that prior Rogers-Young (RY)
closure based calculations of S(q) (and our own HNC-based computations) found a non-
monotonic variation of the cage peak intensity with φ [1]. This is consistent with our real
space results, which to our knowledge have not previously appeared in the literature.
3.3 Naïve MCT Kinetic Arrest Map
Two examples of the dynamic free energy, Feff(r), for many arm stars are shown
in the inset of Fig.3.2. The ideal NMCT crossover is defined as when a barrier in Feff(r)
first emerges. One curve (black) is at a volume fraction just below the ideal NMCT
crossover, the other (red) is at a high value where the barrier is 8B
k T . For hard spheres
no kinetic divergences are predicted at volume fractions below random close packing [9].
36
The NMCT ideal kinetic arrest diagram is then constructed in the main frame of
Fig.3.2. Localization on the star scale is not predicted if the particles are too soft (f < 40),
and non-monotonic behavior occurs at intermediate arm number. The NMCT boundary is
compared with a more elaborate full ideal MCT calculation based on RY closure input,
and also an iso-diffusivity boundary deduced from simulation [10]. Good agreement is
obtained which establishes the important point that NMCT theory with HNC input
reliably captures subtle ideal dynamical arrest phenomena for many arm stars.
3.4 Activated Barrier Hopping and Dynamical Fragility
The NLE theory is now employed to perform the study of activated barrier
hopping and dynamical fragility in glassy star polymer fluids.
Fig.3.3 presents the dynamic free energy barrier height as a function of volume
fraction. The barrier height increases roughly linearly with arm number below J
φ . Once
beyondJ
φ , the barrier saturates. This behavior dominates the mean hopping time trends,
which is similar shown in log-linear format. Fig.3.4 shows the dimensionless mean
barrier hopping time,hop
τ , which grows strongly with both volume fraction and arm
number below the soft jamming threshold. The increase is stronger than Arrhenius
defined as an exponential growth, exp( )hop
bτ φ∝ . The predicted supra-Arrhenius
behavior can be tested via relaxation, viscosity, and/or self-diffusion experiments, which
are beginning to be performed [11]. BeyondJ
φ , the hopping time saturates.
How tunable star softness impacts “dynamic fragility” below the jamming
crossover as kinetic vitrification is approached is a fundamental issue of broad interest.
This question is framed in analogy with the “Angell plot” employed for thermal glass
37
formers [12] where the relaxation time of different systems is normalized to a common
point at the kinetic glass transition. For many arm stars, we adopt the natural fragility
definition employed in recent experiment [13] and simulation [14] studies of microgel
suspensions: (log( / ) / ( / ) |g
hop s gmφ φ φτ τ φ φ =≡ ∂ ∂ . The glass transition volume fraction,
gφ ,
follows from the kinetic criterion / 10y
hop sτ τ = . In colloid experiments and simulations
[13,14] a typical y ~ 4. The dynamic fragility results using y=4 are shown in Fig.3.5. A
remarkably wide range of dynamic fragilities is predicted, increasing from ~ 8 for f=150
where the relaxation is effectively Arrhenius and the soft stars behave as a “strong glass
former” as observed in recent microgel experiments [13], to ~ 43 in the hard sphere limit.
This roughly factor of five fragility variation is almost as large as seen in all thermal glass
formers (metals, molecules, network formers) [12]. Physically, it is a dynamical
consequence of the enhanced sensitivity of local structure (which determines the barrier)
to φ as individual star polymers interact via a repulsion that is spatially more rapidly
varying (higher f). For experimentally relevant intermediate values of f, the inset of
Fig.3.6 shows mφ is predicted to increase logarithmically with arm number. The
theoretical results are consistent with the limited experimental viscosity data available
which do display a monotonic increase of fragility with increasing arm number [15,16].
The above dynamic fragility trends are not sensitive to the adopted vitrification
criterion, as indicated by the y=9 results in the inset of Fig.3.5. One sees the dynamic
fragility again varies by a factor of ~5-6 (hard sphere 83mφ ). Kinetic glass formation
boundaries based on the two vitrification criteria are presented in Fig.3.2. Modest shifts
towards larger volume fractions relative to the NMCT ideal glass boundary are found, but
the gross features of the dynamical crossover and kinetic arrest boundaries agree.
38
3.5 Linear Elastic Shear Modulus and Experiment-Theory Confrontations
Numerical calculations of the linear elastic modulus are shown in Fig.3.6 for an
experimentally relevant range of star arm numbers. All soft colloids display a vastly
weaker response to volume fraction than hard spheres ( f → ∞ ). As arm number
increases, the modulus monotonically grows, and a rough power law volume fraction
dependence emerges with an effective exponent that increases linearly with arm number
(for 267f ≤ ) as: ' , 2.4 0.01nG n fφ∝ ≅ + . Such power laws are, of course, approximate
in that they apply only over a limited volume fraction regime, as also true in experiment.
We have verified that the hard sphere result is numerically recovered beyond an
experimentally impossible to achieve large value of f >1000 (not shown). Above the soft
jamming threshold, G’ fundamentally changes and exhibits a weak linear growth with
volume fraction for all stars for the same physical reason that the hopping time saturates.
Such a linear increase has been observed in ultra-high volume fraction microgel
suspensions [17].
The theoretical results are compared in Fig.3.6 with an experiment for f=267
polybutadiene stars [15]. The experimental concentration variable (c/c*) is converted to
its theory analog as
( )3
3 3/ 6 (4 / 3) / 2 0.26 / *g g
R R c cφ ρπσ ρ π σ= = = (3.2)
whereg
R is the radius of gyration and h
R is the hydrodynamics radius
and 16g h
R R≈ = nm [1,15]. The experimental shear modulus data [15] in the theory units
has been shifted up by a constant factor of 42. The theory accurately predicts the apparent
39
power law dependence. Very recently, new experiments on f=128 stars have been
performed at volume fractions beyond the soft jamming threshold [18]. The observed φ
dependence of G’ becomes dramatically softer and is nearly linear.
3.6 Ultralocal Analysis
Fig.3.7 shows that FB is tightly correlated with the coupling constant, λ ,
consistent with prior hard sphere analysis [19], a measure of the dynamical mean square
force exerted on a tagged particle due to repulsive force collisions. And for soft particles,
its definition follows from the variation of Eq.(2.10): 2
1gφλ = . The volume fraction
dependence of λ (Fig.3.8) explains the (near) saturation of FB above J
φ .
We have numerically verified that all trends in Fig.3.6 can be well understood
based on the approximate analytic result derived for hard spheres within the NLE
approach: ( )23 2' / ( , )
LG r fβσ φ σ φ φ λ
−∝ ∝ [19]. The theory predicts the gas-like growth
of G’ is a consequence of the (near) saturation of the localization length and coupling
constant (Fig.3.8). Fig.3.9 further demonstrates all volume fraction and arm number
dependences nearly collapse onto a single curve (for 267f ≤ ) as a function of λ both
above and below J
φ . However, the effective slope is close to unity, substantively
different from its hard sphere value of two [19]. Hence, the finite range of interparticle
repulsions modifies the precise connection between elasticity and local structure.
Interestingly, since the barrier and modulus both grow roughly linearly with λ ,
an approximate proportionality between G’ and FB emerges, reminiscent of the “elastic
shoving model” of thermal glass formers [20]. Hence, a direct and deep connection
40
between the barrier, which controls long time relaxation, and the glassy modulus, which
controls short time elasticity, is predicted.
3.7 Summary and Discussion
In this chapter, we have presented the first theoretical study of the glassy behavior
of dense fluids of many arm star polymers that addresses in a unified manner real space
structure, elasticity, and relaxation via activated barrier hopping. A “thermal vestige” [8]
of jamming is identified structurally which has massive viscoelastic consequences. Below
the soft jamming crossover, the shear modulus follows a f-dependent power law
concentration scaling, relaxation is non-Arrhenius, and fragility increases logarithmically
with arm number. An exceptionally wide range of fragilities are predicted based on
activated hopping as the dominant relaxation mechanism. It is relevant to note that a
recent application of ideal MCT to fluids that interact via a tunably soft finite range
harmonic repulsion (model microgel) find MCT qualitatively fails to predict the large
change of fragility seen in simulations [6]. This reinforces a conclusion of recent
theoretical [19,21,22], simulation [23,24], and experimental [25] work on hard spheres
that activated hopping is very important at experimentally accessible high volume
fractions. Above the crossover volume fraction J
φ , the shear modulus is predicted to
weakly increase in a linear manner, while the relaxation time almost saturates.
One caveat involves the role of chain entanglements between arms on two
interpenetrating stars at very high volume fractions [11,15] which is not accounted for by
the present (or any other) theory. We believe this explicitly polymeric effect is not an
important limitation for the highly localized physics that determines the glassy elastic
41
modulus. However, the slow relaxation process will be influenced, and our predicted
saturation of the hopping time beyond J
φ likely does not occur if the star arm length
exceeds the entanglement threshold. On the other hand, direct simulations of the model
defined by Eq.(3.1) can test our results at ultra-high volume fractions. Moreover, it would
be interesting to perform new experiments with stars of arm length below the
entanglement threshold. Finally, we emphasize the generality of our NLE approach
which can be employed to study other soft repulsive or attractive colloids, and to explore
the role of particle softness on dynamic heterogeneity [22] and nonlinear rheology [26].
3.8 References
[1] Likos C.N., Phys.Rep., 348 (2001), 267; Jusufi A., Watzlawek M. and Lowen H.,
Macromolecules, 32, (1999), 4470.
[2] Stellbrink J. et al., Progr. Colloid Polym. Sci. 115 (2000), 88.
[3] Witten T.A. and Pincus P.A., Macromolecules, 19 (1986), 2509.
[4] Hansen J.P. and McDonald I.R., Theory of Simple Liquids (Academic Press, London,
1986).
[5] Although the computationally more intensive Rogers-Young closure is quantitatively
more accurate for many arm stars [1], we employ the HNC closure for numerical
convenience and because it allows comparison with our ongoing studies (and those of
others [6,7]) of model microgel suspensions based on harmonic repulsive potentials
which employ this closure. As discussed in the text, and elsewhere [7], the HNC closure
is reliable for the questions addressed.
[6] Berthier L., Flenner E., Jacquin H. and Szamel G., Phys.Rev.E, 81 (2010), 031505.
42
[7] Jacquin H. and Berthier L., cond-mat: arXiv: 0912.2657 (2009).
[8] Zhang Z., Xu N., Chen D., Yunker P., Alsayed A.M.,Aptowicz K.B., Habdas P., Liu
A.J., Nagel S.R. and Yodh A.G., Nature, 459 (2009), 230.
[9] Schweizer K.S. and Yatsenko G., J.Chem.Phys.,127 (2007), 164505.
[10] Foffi G., Sciortino F., Tartaglia P., Zaccarelli E., LoVerso F., Reatto L., Dawson
K.A. and Likos C.N., Phys. Rev. Lett., 90 (2003), 238301.
[11] Erwin B.M,Cloitre M.,Gauthier M. and Vlassopoulos D.,Soft Matter,submitted,
2010.
[12] Angell C.A., Ngai K.L., McKenna G.B., McMillan P.F. and Martin S.W., J.App.
Phys. 88 (2000), 3113.
[13] Mattson J., Wyss H.M., Fernando-Nieves A., Miyazaki K., Hu Z., Reichman D.R.
and Weitz D.A., Nature, 462 (2009), 83.
[14] Berthier L. and Witten T.A., Phys.Rev.E, 80 (2009), 021502
[15] Ozon F., Petekidis G. and Vlassopoulos D., Ind. Eng. Chem. Res.,45 (2006),6946.
[16] Vlassopoulos D., Fytas G., Pispas S. and Hadjichristidis N., Physica B, 296,(2001),
184.
[17] Seth J.R., Cloitre M. and Bonnecaze R.T., J.Rheology, 50 (2006), 353.
[18] Erwin B.M,Cloitre M.,Gauthier M. and Vlassopoulos D.,Soft Matter,6, 2825, 2010.
[19] Schweizer K.S. and Yatsenko G., J.Chem.Phys.,127 (2007), 164505.
[20] Dyre J., Rev.Mod.Phys.,78 (2006), 953.
[21] Schweizer K.S. and Saltzman E.J., J.Chem.Phys.,119 (2003), 1181; Schweizer K.S.,
J.Chem.Phys.,123(2005),244501;SchweizerK.S.,Curr.Opin.Coll.Inter.Sci.,12(2007), 297.
43
[22] Saltzman E.J. and Schweizer K.S., Phys.Rev.E, 74 (2006), 061501 and 77 (2008),
051504.
[23] Kumar S.K., Szamel G. and Douglas J.F., J.Chem.Phys.,124 (2006), 214501;
Brumer Y. and Reichman D.R., Phys.Rev.E, 69(2004), 041202.
[24] Brambilla G., ElMasri D., Pierno M., Berthier L., Cipelletti L., Petekidis G. and
Schofield A.B., Phys.Rev.Lett.,102 (2009), 085703.
[25] Brambilla G., ElMasri D., Pierno M., Berthier L., Cipelletti L., Petekidis G. and
Schofield A.B., Phys.Rev.Lett.,102 (2009), 085703.
[26] Kobelev V. and Schweizer K.S., Phys.Rev.E, 71 (2005), 021401.
44
3.9 Figures
Fig. 3. 1
Radial distribution function for f=128 stars at several volume fractions. Inset: Primary
cage peak (upper black curve) and inner peak (lower red curve) heights as a function of
volume fraction.
45
Fig. 3. 2
Comparison of NMCT ideal glass boundary (green) with a simulation iso-diffusivity
curve (black; D=0.017) and full ideal MCT calculation (red) [16]. NLE theory kinetic
glass boundaries for two vitrification criteria are also shown (blue : 4/ 10hop s
τ τ = ;
magenta: 9/ 10hop s
τ τ = ). Inset: dynamic free energy for f=128 stars at a volume fraction
just below the ideal NMCT crossover (black) and a high value where the barrier is 8 kBT
(red).
46
Fig. 3. 3
Dimensionless dynamic barrier height is shown as a function of volume fraction for
(from bottom to top): f= 56, 64, 96, 128, 200, 267.
47
Fig. 3. 4
Dimensionless mean barrier hopping time is shown as a function of volume fraction for
various arm numbers.
48
Fig. 3. 5
Fragility plot of the mean hopping time versus normalized volume fraction for a kinetic
glass criterion of 4/ 10hop s
τ τ = and arm numbers (from bottom to top): hard sphere, 1000,
500, 350, 267, 200, 150. Inset: fragility as a logarithmic function of arm number for two
kinetic vitrification criteria: 4/ 10hop s
τ τ = , 9/ 10hop s
τ τ = .
49
Fig. 3. 6
Dimensionless shear modulus, * 3 'G Gβσ≡ , as a function of volume fraction for various
star arm numbers including the hard sphere limit (black curve). Open squares are shifted
[27] experimental data [7] for f=267.
50
Fig. 3. 7
Dynamic barrier height is shown as a function of coupling constant for f=128, 200, 267.
HS calculation is also shown for comparison.
51
Fig. 3. 8
Coupling constant is shown as a function of volume fraction for f=128, 200, 267.
52
Fig. 3. 9
* /G φ is shown as a function of the coupling constant. The hard sphere (black squares)
calculation is also shown for comparison.
53
CHAPTER 4
GLASSY DYNAMICS AND MECHANICAL RESPONSE IN
DENSE FLUIDS OF HERTZIAN SPHERES
4.1 Introduction
This chapter presents applications of the methodology developed in Chapter 2 to
dense suspensions of microgels modeled as repulsive Hertzian spheres at the center of
mass level. Microgels are intramolecular crosslinked polymeric networks swollen by
solvent with diameters that span the range from ~10 nm to a micron or larger. They are
largely impenetrable, and have been widely studied due to their importance as advanced
functional materials [1]. Microgel single particle softness or elasticity is highly variable
via manipulation of solvent quality and crosslink density.
Recent computer simulation studies of the quiescent dynamics and elasticity of
the Herztian contact and closely related (so-called parabolic potential
( )2
( ) 1 ( / ) ,V r r rε σ σ= − ≤ ) finite range soft repulsion models have observed some of
the six features discussed in the introduction of Chapter 2 [1-3]. An interesting two-
branch universality of the temperature and concentration dependences of the mean alpha
relaxation time has been uncovered [3]. More generally, recent theoretical and simulation
studies have also discovered a rich equilibrium phase behavior, including re-entrant phase
transitions at ultra-high volume fractions due to subtle competition of energetic and
entropic factors associated with soft particle overlap [4].
The remainder of this chapter is structured as follows. Section 4.2 discusses
interaction potential, equilibrium pair correlations and “soft jamming” crossover. Naïve
54
MCT calculations of the ideal glass transition are also performed of the dynamic
crossover “phase diagram”. The key features of the dynamic free energy that determines
activated relaxation are studied in section 4.3 as a function of volume fraction and
particle softness. The influence of the latter two variables on the mean relaxation time is
established in section 4.4, and the question of a master curve construction explored.
Section 4.5 presents the kinetic arrest diagram, and analyzes how dynamic fragility
depends on particle softness. Calculations of the quiescent shear modulus and localization
length as a function of volume fraction and particle stiffness or repulsion strength are
presented in section 4.6. Remarkable connections between the shear modulus and
activated relaxation process are established in section 4.7, including a microscopic basis
for the so-called elastic “shoving model” [5] of highly viscous liquid relaxation, and the
recent experimental observation of a direct relation between elasticity and kinetic glass
formation in microgel suspensions [6]. Section 4.8 studies the yield stress and strain,
shear thinning of the viscosity and relaxation time, and stress-strain rate flow curves. The
chapter concludes in section 4.9 with a summary and discussion. Section 4.10 and 4.11
present the references and figures, respectively.
4.2 Model, Equilibrium Structure and Dynamic Crossover
4.2.1 Hertzian model
The classic soft repulsive Herztian pair potential is:
5/ 24( ) (1 / ) ,
15
( ) 0,
V r E r r
V r r
β σ σ
β σ
= − ≤
= >
(4.1)
55
where r is interparticle separation, σ the particle diameter, 1
Bk T β −=
the thermal energy,
and 4E/15 is an inverse dimensionless temperature which quantifies repulsion softness or
particle stiffness [7]. For microgels, * 3
BE E k Tσ −= is a tunable single particle modulus
controlled largely by internal polymer crosslink density. Eq.(4.1) describes as single
contact, and is believed to be realistic within a pair decomposable potential energy
description when particle deformation is of order 10% or less [1]. The hard sphere
potential is smoothly recovered as E → ∞ . Fluid volume fraction is 3 / 6φ ρπσ= , where
ρ is the particle number density. This pair potential is an oversimplification for
crosslinked polymer microgels, and cannot be expected to be reliable at very high volume
fraction where Eq.(4.1) allows significant sphere overlaps, in contrast to real systems
which deform, facet, and osmotically deswell [1].
4.2.2 Equilibrium structure and soft jamming crossover
Fig.4.1 shows representative results for the pair correlation function as a function
of volume fraction at fixed E. With increasing concentration, the primary cage peak (r~σ
initially increases and very weakly shifts to smaller separations. However, beyond a
characteristic volume fraction of 0.8 0.9J
φ φ≅ − ≡ (generally increases as E decreases),
the primary ordering peak decreases with concentration, and the second peak splits. We
refer to J
φ as a soft jamming threshold in analogy with a recent experimental (simulation)
study of microgel (soft repulsive) fluids [8] which found the primary peak amplitude, 1g ,
goes through a maximum, an effect called the “thermal vestige of jamming” [9]. We also
note the OZ-HNC theory predicts a split second peak in g(r) at high volume fractions, as
often seen in glasses and microgel experiments [8].
56
The inset of Fig.4.1 presents the corresponding static structure factors, the most
obvious features of which monotonically vary with volume fraction. Hence, these soft
particles present an interesting situation where the dependence on volume fraction of the
two classic measures of short range order, the cage peak in g(r) and S(k), do not vary the
same with concentration at high volume fractions. This reflects subtle entropic and
energetic effects when particles significantly overlap.
Fig.4.2 plots the cage peak maximum of g(r), 1g , as a function of volume fraction
for a wide range of E values spanning the range of ultra-soft to near hard sphere.
Motivated by the work in [8], a differential volume fraction variable is employed:
Cφ φ φ∆ = − , where
Cφ is the hard sphere RCP value. For internal consistency, we
employ our best numerical estimate of C
φ based on the HNC integral equation theory.
The precise value is not analytically known, and we have extracted it (see inset) by
analyzing the dimensionless isothermal compressibility, 0( 0)S q S= ≡ . Numerically
converged results of the OZ-HNC integral equation are obtained up to 0.75φ = , the
quantity 0S∆ is then plotted versus φ in the highest volume fraction regime of (from 0.6
to 0.75) corresponding to an order of magnitude variation of S0. The exponent is varied to
best linearize the plot resulting in 3/ 4∆ ∼ . Modest linear extrapolation to the
incompressible limit then yields an apparent jammed state at 0.78J
φ . This value is well
above the correct RCP result of ~ 0.64, which is not surprising given the known fact that
RCP for the PY closure equals unity and hence incorrect [10].
Each curve for different E values in Fig.4.2 exhibits a maximum, except those
very close to hard spheres. For the latter, reliable numerical results beyond φ ~0.75 could
57
not be obtained due to convergence problems of the OZ-HNC integral equation. Note that
1g intensifies and shifts to higher volume fractions as particles stiffen. This feature is
relevant to the rich variation of the dynamics with particle stiffness discussed later.
From Fig.4.2 one sees that J
φ is a weakly non-monotonic function of E (see also
Fig.4.3). However, this is a subtle point which we view as a minor quantitative effect that
occurs under ultra-soft repulsion conditions where 1( )g φ is a low intensity, broad function.
Hence, we only wish to emphasize that the soft jamming crossover varies very slowly as
E becomes small. Note that although literal “contacts” in the sense defined for hard
spheres are not present, integration under the first peak of g(r) yields the classic measure
of number of nearest neighbors. Interestingly, we find this number is almost universal at
Jφ and equal to ~12-13, which is physically sensible.
All the above structural trends appear to agree well with recent simulations
[8,11,12] and experiments [8], which provides support for the zeroth order reliability of
the OZ-HNC static correlations as input to dynamical theories.
4.2.3 Naïve MCT kinetic arrest map
Fig.4.3 shows the NMCT kinetic arrest or dynamic crossover map; the soft
jamming structural crossover discussed in section 4.2.2 is also indicated. As E → ∞ the
hard sphere ideal glass transition of NMCT (based on HNC input) at 0.445MCT
φ = is
recovered. No ideal localization transition is predicted if the particles are sufficiently
soft, 600E < . A physical interpretation is that at low temperatures (high E, stiff particles)
the dynamic caging constraints are strong and upon increasing volume fraction the
system reaches the glass transition before the soft jamming crossover beyond which
caging constraints weaken due to local structural disordering. However, for sufficiently
58
high temperatures (soft particles), the system reaches the soft jamming threshold before
caging constraints become strong enough to result in localization. Interestingly, the
functional form of the NMCT ideal nonergodicity boundary is well described as a
parabolic critical power law: ( )21
c c MCTT E φ φ−∝ ∝ − , in agreement with the recent full
MCT [11] and simulation [3,11] studies.
The full ideal MCT with OZ-HNC input has very recently been applied to the
Herztian contact model up to ultra-high volume fractions [13]. The effective temperature
(~1/E) of ideal kinetic arrest was found to be a non-monotonic function of volume
fraction with a maximum near φ ~1. This behavior was termed “anomalous”, and agrees
with empirical fits of MCT formulas to the relaxation time simulation data in the
crossover regime. In the inset of Fig.4.3 we present the analogous simple NMCT result.
Indeed, NMCT also predicts re-entrance at extremely high volume fractions, also at φ ~1,
in excellent accord with the full MCT calculations [11]. The underlying physics relates to
the overlapping of soft particles at very high volume fractions and the corresponding
local disordering of the cage (Figs.4.1 and 4.2).
4.2.4 Dynamic free energy profile
An example Feff(r) is shown in Fig.4.4 for a fixed value of E at three volume
fractions: just below the NMCT crossover, and when the barrier is ~ 5 kBT and 10 kBT.
The hard sphere result at a volume fraction adjusted so that the barrier is 10 kBT is also
shown; the agreement with the analogous soft particle result is excellent. This provides
empirical support that the prior analytic ultra-local analysis of NLE theory for hard
spheres [14,15], which predicts when barriers are high all features of the dynamic free
59
energy are controlled, to leading order, by a single “coupling constant” defined by local
structure in g(r), remains relevant for soft repulsive particles.
4.3 Effect of Particle Softness on Dynamic Free Energy
4.3.1 Localization and barrier lengths
Two key length scales of the dynamic free energy are the (transient) localization
length (rloc) and barrier location (rB). Fig.4.5 shows results as a function of volume
fraction for three values of repulsion strength (E) and the hard sphere limit. The values of
E chosen are typical of estimates of single microgel rigidity of some experimental
systems based on crosslinked polymer network elasticity theory [1]. The qualitative
dependence of the two length scales on volume fraction is similar for all four systems.
For soft particles, the barrier location exhibits a 2-step like growth with concentration,
and then goes through a shallow maximum at a volume fraction a little beyond unity. At
fixed volume fraction, rB grows monotonically with particle stiffness (E). The
corresponding localization lengths initially decrease strongly with φ , but then tend to
saturate, or even very weakly increase, at ultra-high concentrations. The weak non-
monotonic behavior correlates with the non-monotonic variation of local structure and
the soft jamming crossover (Figs. 4.1 and 4.2). As discussed later, the localization length
is the key quantity that determines the elastic shear modulus.
4.3.2 Barrier Height
Barrier heights as a function of volume fraction over a wide range of repulsive
interaction strengths are shown in Fig.4.6. As expected, the barrier height grows
monotonically with E at fixed volume fraction. For all soft particle systems shown, the
60
barrier initially increases in a nonlinear manner with φ , and more quickly for larger E,
reflecting the enhanced sensitivity of local structure to volume fraction for stiffer
particles or equivalently, lower temperature (Fig.4.2). As the soft jamming crossover is
approached, the barrier goes through a maximum, analogous to how the intensity of the
local cage peak of g(r) changes. This connection is a natural one based on the ultra-local
analytic analysis and Eq.(2.11). An interesting consequence of the non-monotonic
behavior is that if particles are sufficiently soft, then a well-defined activated dynamical
regime will not occur since the maximum barrier height never significantly exceeds
thermal energy. For hard sphere limit, the barrier grows monotonically and diverges at
RCP limit as:
( )22 ( )
B HS RCPF g σ φ φ
−∝ ∝ − (4.2)
corresponding to a double essential singularity of the mean hopping time [14,15].
4.3.3 Scaling of barrier with the coupling constant
The barrier height calculations are suggestive that the ultra-local analysis result
for hard core particles that 2
1gλ φ= controls the barrier height may be accurate for
Herztian repulsive particles. However, this does not mean Eq.(2.11) literally applies since
the linear connection between the barrier and coupling constant emerges as a
consequence of the specific hard sphere high wavevector behavior of the dynamical
vertex [14]. We examine this issue in Fig.4.7, where the barrier heights for three soft
particles are plotted versus the coupling constant. Over a significant range of FB, the soft
particle results do roughly collapse for volume fractions below the soft jamming
threshold. However, for soft particles with intermediate values of E (where results for
λ >20-30 cannot be attained given the soft jamming structural behavior), the slope is not
61
unity as in Eq.(2.11) for hard spheres (and also many arm stars [16]), but rather
roughly 2
BF λ∝ . On the other hand, Fig.4.7 also shows that for the highest E Hertzian
system, a crossover from 2
BF λ∝ scaling to something approaching the linear scaling of
hard spheres occurs at very large barriers (highest volume fractions). This change for stiff
particles must occur given the Herztian contact model approaches the hard sphere
potential as E gets very large.
The origin of the differences in the effective scaling of the barrier with the
coupling constant for intermediate stiffness and very stiff (hard sphere like) systems must
be related to the wavevector dependence of the dynamical vertex. Fig.4.8 presents
representative vertex calculations for two soft particles and the hard sphere, where
volume fraction for each system is chosen so they all have the same barrier height of 10
kBT. The hard sphere result (bottom panel) shows the expected saturation of the
amplitude at large k. But, for soft repulsive particles the vertex acquires a maximum at a
finite wavevector, and then decays to zero at large k. The maximum shifts to smaller k as
the particles become less repulsive (lower E), even though the volume fractions have
been adjusted to main the same FB. However, the maximum amplitude is not significantly
dependent on E over the range studied.
We conclude that the qualitative difference of the large wavevector dependence of
the dynamical vertex (very local force correlations) for soft particles of intermediate
stiffness (compared to very stiff hard sphere like cases) is the origin of the different
scaling of the barrier with coupling constant. However, the ability of the coupling
constant to (roughly) collapse the barrier height calculations for soft particles of different
intermediate levels of stiffness is maintained. This rather remarkable, but empirically-
62
deduced, result can perhaps be rationalized based on the near constancy of the maximum
vertex amplitude.
4.4 Mean Alpha Relaxation Time
Using Eq.(2.8) the mean barrier hopping time is computed as a function of
particle softness and volume fraction. All times are expressed in terms
of, 2 2/( ) /s s B s
k T Dτ σ ζ σ= = , where s
D is the measurable “short time” diffusion constant.
4.4.1 Volume fraction and repulsion strength dependences
Fig.4.9 presents the dimensionless mean hopping times as a function of volume
fraction for a wide range of particle stiffness (E). Results are shown over an exceptionally
wide range of relaxation times (20 orders of magnitude) to emphasize the robust nature of
the non-monotonic behavior, as expected from the barrier height calculations in Fig.4.6.
In real colloidal suspensions (and often simulations too), a relaxation time range of only
typically 4-6 orders of magnitude can be probed, perhaps up to 4 6/ ~ 10 10hop s
τ τ − which
represents a practical threshold used to define kinetic vitrification. Fig.4.9 shows the
relaxation times generally grow in a “non-Arrhenius” (nonexponential in φ ) manner
which is stronger as E increases; such nonexponential growth has been observed in
experiment [8,17]. But in all cases the dependence is weaker than the limiting hard
sphere behavior. The particle stiffness (inverse temperature) of the mean relaxation time
is presented in Fig.4.10, now over only a range of times typically relevant to experiments
and simulations (~ five orders of magnitudes). Two qualitatively distinct behaviors are
predicted: (1) asymptotic saturation at the hard sphere value as E → ∞ for the lower
range of volume fractions (roughly below the hard sphere RCP), and (2) increasingly
63
strong growth with E for the highest φ values. These trends are in qualitative accord with
recent simulations [3].
We have briefly explored the possibility of interpreting the relaxation time
calculations in terms of an effective hard sphere reference system. Specifically, an
effective hard core diameter (or equivalently, effective volume fraction,eff
φ ) is defined by
via a dynamical mapping criterion :
( ) ( , )HS soft sphere
hop eff hopEτ φ τ φ= (4.3)
Using Eq.(4.3) and our hopping time calculations, we have extracted the effective volume
fractions for several choices of the maximum value (kinetic glass point) of /hop s
τ τ
ranging from 2 710 10→ . Results are shown in Fig.4.11 for values of the latter relevant
typical colloid experiments or computer simulations, /hop s
τ τ = 2 410 10→ . For these
cases the effective volume fractions are very well described by (T=15/4E):
( , ) b
effT aTφ φ φ≈ − (4.4)
where eff
φ φ< , 0a > , and 0.66b .
Interestingly, our results agree rather well with the form deduced from
simulations or the parabolic finite range repulsion model, and rationalized by specific
physical arguments [3]. However, if the maximum value of /hop s
τ τ is the very high, e.g.,
7/ 10hop s
τ τ → , then we find Eq.(4.4) is again accurate but b~ 0.84 (not plotted). Hence,
based on our dynamical mapping the apparent exponent b depends on the relaxation time
range analyzed. This is perhaps not a surprising result, which may also be influenced by
the accuracy of the OZ-HNC structural input which is likely not uniform as a function of
volume fraction.
64
4.4.2 Universal collapse
Simulations of the finite range parabolic repulsion model fluid has recently
discovered a collapse of the alpha relaxation time ( ατ ) of soft repulsive spheres as a
function of T and φ based on two distinct scaling functions [3]. Physically-motivated
(but not derived) jamming and effective hard sphere arguments which we do not repeat
here suggest the form of the two master curves are:
( ) 2/
0 0exp | | | | /T A F Tδ µ
ατ φ φ φ φ−
±− −∼ (4.5)
where A is a numerical constant, and ( )F x± apply to volume fractions above/below a
critical volume fraction, 0φ , suggested to be the hard sphere RCP value. The exponents µ
and δ were varied with best fit values of 1.3µ ∼ and 2.2δ ∼ obtained, which were
argued to have a theoretical basis related to the specific finite range repulsive potential
[3]. The crossover functions ( )F x± are constrained by recovering the hard sphere limit
and continuity at 0φ φ= : ( ) 1F x− → ∞ → , ( )F x+ → ∞ → ∞ , and / 2( 0)F x xδµ
± → → . All
simulation data then empirically collapse onto the two distinct curves by plotting
0| | log( )Tδ
αφ φ τ− versus 2/
0| | /Tµφ φ− over about 5 orders of magnitude of the
ordinate. Note that in our notation, the temperature scale factor 1/ 2 /s
T E τ∝ .
We analyze our theoretical calculations in Fig.4.9 and 4.10 in an analogous
empirical manner ( 1T E
−≡ ) by freely floated the parameters 0φ , µ and δ to achieve an
optimized master plot. Our primary goal is only to check if the NLE theory is consistent
with such a two master curve description based on the dominance of activated hopping
dynamics. The results are shown in Fig.4.12. We note that the abscissa and ordinate
values and range are quite similar (but a bit narrower) to that probed in simulation [3]. An
65
excellent collapse onto two master curves is achieved using µ = 1.3, 2.2δ = and
0 0.78φ = . Remarkably, these δ and µ are identical to that deduced from simulation [27].
Moreover, the deduced value of 0 0.78φ = agrees with our independent best estimate of
RCP based on the OZ-HNC structural theory.
Now, the functional form of the lower scaling function in Fig.4.12 is very similar
to that obtained in simulation [3], including saturation at large values of the abscissa. We
argue this is the regime the dynamical theory should work best since the inaccurate
description of jamming of the equilibrium OZ-HNC approach will have the least
significance. However, the upper branch in Fig.4.12 (beyond hard sphere RCP) does not
show as good a collapse, and does not appears to be a divergent form on the present scale
as seen in the simulation (i.e., our results tend to “bend over”, versus the simulation
which “curves up”). We do not have a deep understanding of this deviation, but suspect it
is largely, or completely, due to the poor quantitative reliability of the OZ-HNC theory at
very high volume fractions. This interpretation can be tested by using simulation results
for the structural input in NLE theory. We also suspect that the apparent divergent form
of the scaling function at ultra-high volume fractions found in the simulation is only an
apparent or crossover effect, and must eventually “bend over” given the lack of a true
RCP for soft particles and the finite energy cost to fully overlap particles.
Finally, we note that the full ideal MCT analysis of this system predicts very well
(albeit over a tiny volume fraction range of 0.515< φ <0.53) the scaling behavior in
Eq.(4.5) but based on plotting ατ , not log ατ as in the simulation analysis and in Fig.4.12
[11]. The latter reflects the incorrect MCT prediction of a critical power law divergence
of the relaxation time, versus the correct strongly exponential (essential singularity) form
66
observed in the simulation and also NLE theory. Moreover, MCT makes qualitatively
incorrect predictions for the variation of dynamic fragility [11].
4.5 Kinetic Arrest Map and Dynamic Fragility
Knowledge of the mean activated relaxation time allows two additional issues of
significant scientific and practical interest to be addressed: (1) the kinetic vitrification
map, and (2) the dependence of a dynamic fragility on single particle stiffness.
4.5.1 Vitrification boundaries
Kinetic vitrification is defined via a maximum relaxation time criterion.
Interesting questions include how it depends on the criterion adopted, and how it
compares with the dynamic crossover (ideal MCT) boundary. Calculations for three
experimentally-relevant criteria are shown in Fig.4.13. Modest shifts towards larger
volume fractions of the kinetic vitrification volume fraction, g
φ , are predicted as the time
scale for kinetic arrest grows. Hence, the threshold E (or temperature) required to clearly
observe activated dynamics is significantly higher (or lower) than the MCT crossover
value. However, the shape of the kinetic arrest boundaries are well described by the same
parabolic form found for the NMCT analog, albeit with significantly larger apparent
(empirically fit) values of C
φ . These findings are consistent with simulation [3].
Moreover, as discussed in previous contexts, they suggest a close correlation of the cage
physics that enters MCT and NLE theory [18-21].
We also note that the kinetic arrest and soft jamming crossover boundaries cross
when particles become soft enough (high effective temperature). This crossing, and the
67
lack of any deep connection between kinetic vitrification and soft jamming, is consistent
with conclusions drawn from recent simulations [8].
4.5.2 Dynamic fragility
Given knowledge of the kinetic vitrification volume fraction (analog of Tg for
thermal glass formers), one can determine how repulsion softness impacts dynamic
fragility. We analyze this question below the soft jamming crossover using the results in
Fig.4.9 and 4.13, and adopt a perspective analogous to the classic Angell plot employed
for thermal glass formers. Specifically, fragility is defined per recent experiments as [17]:
(log( / ) / ( / ) |g
hop s gmφ φ φτ τ φ φ =≡ ∂ ∂ , (4.6)
where g
φ follows from the criterion / 10y
hop sτ τ = . In experiments [8,17,22,23] and
simulations [3,22,24], a typical y ~ 2-5. A fragility plot using y=3 is shown in Fig.4.14
based on a relaxation time variation of 4 orders magnitude as is typical of colloid
experiments. A wide range of dynamic fragilities is predicted. For example, mφ ~18 for
the rather low value of E=6000 where the relaxation is effectively “Arrhenius”
corresponding to “strong glass” behavior. This is qualitatively consistent with the recent
experimental discovery that “soft microgels make strong glasses” [17]. With decreasing
temperature, fragility monotonically grows, approaching mφ ~41 in the extreme fragile
hard sphere limit. The growth of fragility with particle stiffness reflects the enhanced
sensitivity of local structure to changes of volume fraction (Figs. 4.1 and 4.2). We also
predict the dynamic fragility decreases linearly with increasing temperature (Fig.4.14
inset). Moreover, one sees that the overall trends are not very sensitive to the vitrification
68
criterion although the quantitative values of mφ are. Future simulation or experiments
should be carried out to test these findings.
4.6 Linear Shear Modulus
Fig.4.15 presents calculations of the linear shear modulus, in units of 3
Bk Tσ − , as a
function of volume fraction over a wide range of dimensionless repulsion strength, E, that
quantifies the Hertzian repulsion. All soft particle systems display a much weaker
response to volume fraction than hard spheres. The present theory quite accurately
accounts for the volume fraction dependence of the glassy shear modulus of hard sphere
suspensions [25,26]. Quantitatively, a dimensionless shear modulus of ~10-104
corresponds to ~ 40-40,000 Pa for a σ =100 nm diameter particle at room temperature.
The growth of G’ with volume fraction does not display any simple functional form over
a wide range. Over a narrow intermediate range of volume fractions, apparent power law
growth, ' ~ nG φ , is present for sufficiently stiff particles, where n ~ 7-9. At high
concentrations, the modulus decreases because of the reduction of local packing order
associated with the soft jamming crossover.
We have numerically verified that Eq.(2.12) relating G’, φ and the localization
length is well obeyed. The inset of Fig.4.16 presents some representative evidence for
this statement; ones sees that over a range where the dimensionless modulus grows by a
factor of ~30-40, deviations from Eq.(2.12) are a factor of ~1.1-1.5. This theoretical
result agrees with combined confocal and mechanical measurements on microgel
suspensions which reveal “perfect agreement” between the low frequency macroscopic
69
glassy shear modulus and the local “cage modulus” estimated from a relation like
Eq.(2.12) [27,28].
To make qualitative contact with microgel experiments, and choose model
parameters of practical relevance, we note that three samples in a recent study [29] were
characterized by an average number of monomers between two crosslinks of 140X
N = ,
70 and 28. Adopting ideal polymer network theory [30], the corresponding dimensionless
Hertzian stiffness parameters at room temperature is estimated to be: E 4259= , 5832 and
11718. For these systems, the theory predicts apparent shear modulus power law
exponents of n~8-9, while experiments [29] find exponents of n~6-7. Creep
measurements on dense hard spheres with an attached microgel coating [31] also find the
shear modulus grows with volume fraction far weaker than in hard sphere suspensions,
and displays an apparent power law behavior with an exponent of n~5-7.
However, in real microgel systems beyond a threshold volume fraction (analog of
RCP) the particles generally deswell and facet [1], and a generic linear growth with
concentration of the shear modulus emerges [1,7]. This behavior is not captured by our
present approach. We believe the reason is because the soft repulsive sphere model must
have limitations when particles are forced to significantly overlap. Future computer
simulations that compute G’ of the Hertzian model in the equilibrated (glassy) fluid state
can test our results, especially the non-monotonic behavior in Fig.4.15.
The main frame of Fig.4.16 tests the analytic prediction that the dimensionless
shear modulus divided by the volume fraction scales as the square of the “coupling
constant”, 2
1gλ φ≡ . Recall the latter quantifies the magnitude of caging constraints under
high barrier conditions, where g1 is the primary peak height of the pair correlation
70
function [14]. It appears this connection is reasonably well obeyed, in terms of both the
quadratic power law scaling and as a tool to collapse the modulus data for particles of
different repulsion strengths, at least for the intermediate E values studied in Fig.4.16.
4.7 Connections between the Elastic Modulus, Alpha Relaxation and Kinetic Arrest
Puzzling connections between short length and time dynamical or elastic
properties, and the slow and longer length scale alpha relaxation process, have repeatedly
been found experimentally for diverse highly viscous liquids [5]. For example, the
phenomenological shoving model [5] asserts that the activation barrier and temperature-
dependent glassy elastic shear modulus are linearly correlated [32]:
'( ) ( )B c
F T G T V= (4.7)
where Vc is a material-specific, but weakly thermodynamic state dependent, volume.
There is no first principles theory for this quantity, but fitting of Eq.(4.7) to experiments
on thermal liquids typically yields a value much smaller than the elementary atomic or
molecular volume (Vm); for example, for orthoterphenyl (OTP), Vc ~ Vm /30 based on
modeling OTP as a sphere of diameter 0.7 nm [39]. Eq.(4.7) implies local elastic physics
controls the alpha relaxation process and hence dynamic fragility. Remarkably, such a
relation also appears to be valid for dense suspensions of soft repulsive microgels [17],
where Vc/Vm depends on particle stiffness. In this section we test whether NLE theory
predicts a connection between the shear modulus, alpha relaxation time, and kinetic
vitrification volume fraction for dense fluids of soft repulsive spheres.
4.7.1 Barrier-modulus correlation
71
Fig.4.17 cross plots the theoretical dimensionless barrier computed in section
4.3.2 versus the dimensionless fluid modulus per particle. A reasonably good, but clearly
only approximate, linear relation is predicted for soft particles of significantly different
repulsion strengths:
2
3 1'B B
LOC
F G k Tr
σσ φ −
∝ ∝
(4.8)
The inset shows linearity holds quite well for different repulsion strengths up to barrier
heights of ~10 kBT; significantly, the latter range is relevant to the recent microgel
experiments where barriers were deduced to be ~7 kBT or smaller for intermediate and
soft particles [17]. Quantitatively, the inset of Fig.4.17 implies: 3'( / 6)B
F G bπσ , where
0.02 0.08b ≈ − for the three soft repulsive particles and φ ~0.5-1. Hence, in the language
of the shoving model the theoretical Vc/Vm ratio lies in the range of 2-8%, reasonable
values in comparison with the microgel experiments [17]. Eq.(4.8) is understandable
within the context of our approach from the demonstration in section IIIB that 2
BF λ∝
and the results in Fig.4.16. A corrallory is the proportionality of the barrier with the
inverse transient localization length.
The above behavior for thermal soft repulsive particles stands is in contrast with
hard spheres (as shown in Fig.4.17) where numerical studies and analytic analysis have
established [33]: 1 3( / ) '/L B
r F Gσ βσ φ λ− ∝ ∝ ∝ . This difference is no doubt related to
the singular nature of the hard core repulsion and associated discontinuity of the radial
distribution function at contact which results in a qualitatively different high wavevector
scaling of the dynamical vertex (see Fig.4.8). Interestingly, our recent NLE study of
many arm star colloids found [16]: '/B
F G φ λ∝ ∝ . That is, the barrier and shear modulus
72
are proportional, but both follow a linear, not quadratic, scaling with coupling constant.
Hence, we conclude that the behavior of spherical particles which interact via the finite
range Hertzian repulsion, Yukawa plus logarithmic repulsion for stars, and hard core
repulsion can be all different with regards to the absolute and relative dependence of the
activation barrier and shear modulus on coupling constant.
4.7.2 Modulus-vitrification correlation
If the activation barrier scales linearly with G’, then it is natural to expect the
kinetically-defined vitrification volume fraction will also be linearly related to φ . Indeed,
such a correlation has been empirically noted, in a model independent manner, in
microgel experiments [6]. In Fig.4.18, we cross plot our kinetic volume fraction (g
φ )
results versus the dimensionless shear modulus at g
φ . A good linear relation is obtained.
The inset shows the analogous plot but based on the single particle repulsive potential
strength parameter, E, but now in a log-linear format. The latter is adopted to display the
trend over a wide range of single particle stiffness parameters. If the results are re-plotted
in a linear-linear fashion, then over the modest range of modulus values examined in the
main frame the plot is highly nonlinear (not shown). This difference is not unexpected
since the macroscopic modulus, G’, depends on fluid volume fraction in contrast to the 2-
particle pair potential property, E.
4.8 Nonlinear Rheology
Many experiments on high volume fraction microgel suspensions or pastes find
solid-like behavior under quiescent conditions [1]. However, measurements of the
frequency-dependent loss modulus sometimes reveal a rise at the lowest frequencies
73
probed, suggesting the presence of an ultra-slow structural relaxation process at
frequencies (times) too low (too large) to measure with typical rheometers. Practical
engineering motivations have also led to many nonlinear rheological measurements being
performed on microgel suspensions [1,34-36]. Key questions are yielding (stress and
strain), shear thinning of the viscosity, and the flow curve. In this section we study such
questions under the assumption that homogeneous flow is achieved. We again emphasize
our results are based on the soft sphere model which must have limitations with regards
to microgel pastes at very high (jammed) volume fractions due to osmotic deswelling,
facet formation, and other effects [1]. Hence, our theoretical results are most
unambiguously tested via comparison with Brownian dynamics simulations of Hertzian
spheres under large deformations.
4.8.1 Stress dependence of barrier and absolute yielding
Fig.4.19(A) presents representative calculations of the effect of stress on the
dynamical barrier height. A doubly normalized format is adopted to emphasize the
generic behavior for different repulsion strengths and volume fractions. The barrier
decreases with stress in a nonlinear manner, in contrast to the simple Eyring model linear
decrease [37]. Nondimensionalization of stress by its absolute yield value leads to a good,
but not perfect, collapse. The corresponding absolute yield stresses are shown in
Fig.4.19(B), in both linear (main frame) and logarithmic (inset) formats. As true of the
quiescent shear modulus, the functional dependence on volume fraction is not simple,
though narrow regimes of apparent power law variation at intermediate volume fractions
can be seen. Stress also increases the localization length of the dynamic free energy, and
74
hence decreases the shear modulus (not shown), as discussed previously for hard and
sticky spheres [38,39].
Fig.4.20 presents calculations of the absolute and mixed yield strains (see
Eq.(2.17) and Eq.(2.18)) as a function of volume fraction for the same systems studied in
Fig.4.19. We note that the precise numerical value of the yield strain is sensitive to the
numerical prefactor “a” in Eq.(2.14); the yield strain calculations in Fig.4.20 (based on
a= π /6) change in a linear manner with this parameter. The computed yield strains
display a relatively weak dependence on volume fraction and repulsion strength
compared to the many orders of magnitude variation in the yield stress and modulus. This
reflects their similar dependences on volume fraction. The absolute yield strains initially
grow rapidly (low barrier regime), and then nearly saturate or weakly decrease at very
high concentrations. The mixed yield strain analogs behave similarly, but weakly
increase at high volume fractions. The differences between these two yield strains reflects
the fact that the shear modulus at the absolute yield stress is ~2-3 times smaller than its
quiescent value (a “strain softening” effect, not shown). We have also computed a
“perturbative” yield strain, defined as when the quiescent shear modulus is reduced by 10
%; we find it is ~ 5-9 % for all volume fractions and E values studied in Fig.4.20 (not
shown).
With regards to microgel experiments, yield strains (often defined as our “mixed”
quantity) are typically 5-10%, and weakly dependent on volume fraction and particle
stiffness [1,28,29]. The elastic-solid-like relation in Eq.(2.16), 'y y
Gτ γ= , that our
calculations are based on appears to be well obeyed experimentally [1]. For the microgel-
like system studied via creep and creep recovery measurements [31], the yield strain is a
75
non-monotonic function of volume fraction: ~10% at an effective φ ~0.58, then passes
through a shallow minimum at φ ~0.62, followed by a gentle rise to ~18% at φ ~1.05.
Quantitative comparison of our calculations with these experiments is difficult due to
both the complexities of microgels not captured by the soft sphere model and the
difficulty of precisely knowing the volume fraction.
4.8.2 Shear Thinning and Flow Curves
We now apply Eqs. (2.19) and (2.20) to study the long time dynamical behavior
under constant stress or steady shear conditions. Fig.4.21 presents representative
calculations of the mean barrier hopping time for three repulsion strengths, and the
corresponding hard sphere behavior, as a function of dimensionless stress; the inset
shows the dimensionless hopping times. The quiescent relaxation times vary over 10
orders of magnitude, and all exhibit stress-thinning commencing at a system specific
stress. The main frame nondimensionalizes the hopping time by its quiescent value in
order to probe whether the functional form of the stress-thinning is generic. A reduction
of the relaxation time by an order of magnitude requires a dimensionless stress of ~3–5.
To zeroth order, quite good collapse of the different curves is obtained, although the
curves do “splay apart” at high stresses, as expected.
Calculations of the stress versus shear rate (both nondimensionalized) flow curves
are shown in Fig.4.22. With increasing single particle stiffness (E) and/or volume fraction,
the flow curves are shifted to lower shear rates. For intermediate reduced shear rates, a
power law like regime emerges, ( )s
ντ γτ∝ , the breadth of which grows with increasing
volume fraction or particle stiffness. The apparent exponent, υ , decreases with increasing
volume fraction and/or repulsion strength, varying over the range ~0.08 to 0.38 for the
76
first 7 curves (from left to right) in Fig.4.22. Since experiments or simulations often only
probe a limited shear rate range, such apparent power laws could dominate the
observation window. Also, recall our calculations do not take into account the “bare” or
short time viscoelastic response. This would lead to an upturn of the flow curves at high
shear rates, which may often be the range observed experimentally [1].
The calculations in Fig.4.22 for fixed E =11718 are re-plotted in Fig.4.23 in a
manner motivated by recent experiments that have searched for universal behavior [1,
28,29]. In the experiments the stress is reduced by an absolute yield value, and shear rate
by a time scale formed from the bare relaxation process (“beta” process associated with
the onset of transient localization) and the locally determined glassy shear modulus [29],
/ 's
Gγη , where s
η is the solvent viscosity. Values of / 's
Gγη have been reported to span
the wide range of 10-11
to 100. As a reasonable surrogate of this dimensionless shear rate
we employ: */ ' /s B s
k T G Gγτ ρ γτ≡ . Then, based on a typical value of 0/ ~ 10s
τ τ , we
estimate */s
Gγτ ~100 / 's
Gγη .
Fig.4.23 demonstrates that there is a near collapse of the flow curves in this
representation under high (beyond soft jamming) volume fraction conditions. Moreover,
at intermediate reduced shear rates an apparent power law applies over many orders of
magnitude with a very small slope. We note that the experimental flow curves [1] for
“jammed” microgel pastes can sometimes be characterized as an apparent power law over
roughly 4 orders of magnitude of (low) shear rates with υ ~0.04-0.08. Such experimental
behavior is more often described [1] as indicating a stress plateau, or “yield stress”, of the
empirical Hershel-Buckley form: m
yaτ τ γ= + . The latter form has been observed in
jammed microgel pastes, with an exponent m ~ 0.5, and a universal collapse of [1,29]:
77
1/ 2
1'
s
y
KG
η γτ
τ
= +
(4.9)
We believe the physics underlying the power law in shear rate response in Eq.(4.9) does
not reflect the alpha relaxation process (which the apparent power law in Fig.4.23 does),
but rather a smaller length scale process likely related to elasto-hydrodynamic physics
[1,36] for deformable particles that is absent in our theoretical model. We suspect our
apparent power law prediction with a tiny exponent is more relevant to the small
deviations of experimental flow curves from a flat response at low shear rates, reflecting
the existence of a very slow alpha relaxation.
The theoretical flow curves for different repulsion strengths at high volume
fractions also do not collapse in the representation of Fig.4.23 (not shown), in contrast to
experiment [1,35]. The origin of this disagreement is unclear, but could be a consequence
of real world complexities of microgel particles, our neglect of all dynamical processes
except the slowest alpha relaxation, inaccuracies of the OZ-HNC input to the dynamical
theory at very high volume fractions, and/or nonlocal mechanical effects not present in
the NLE theory which is based on single particle hopping.
Finally, Fig.4.24 presents the shear viscosity normalized by its quiescent value as
a function of the dimensionless Peclet number ( ( 0)hop
Pe γτ τ≡ = ) for a fixed particle
repulsion strength and six volume fractions. At the higher volume fractions beyond the
soft jamming crossover, a good collapse occurs. In all cases, shear thinning begins at
remarkably small values of Pe~0.001-0.01<< 1, as also predicted by NLE theory for
polymer glasses [27]. This behavior reflects the physical idea that flow under strong shear
proceeds via stress-assisted activated barrier hopping, not a literal complete destruction of
78
the barrier. Power law thinning of the viscosity over many orders of magnitude is
generically predicted but with an apparent exponent smaller than unity, roughly 0.8η γ −∝ .
Interestingly, this apparent exponent is very similar to the predictions of NLE theory
[13,40], and the corresponding single particle relaxation experimental observations [41],
for hard sphere colloid suspensions.
4.9 Summary and Discussion
In this chapter, we have presented the first microscopic theoretical study of
activated glassy dynamics in dense fluids of finite range soft repulsive particles. Our
results provide a physical basis for several of the striking characteristic phenomena
described in the Introduction. The alpha relaxation time in the activated hopping regime
is a rich function of volume fraction and temperature, including exhibiting a maximum
value at ultra-high volume fraction due to a soft jamming crossover that signals local
packing disorder due to particle overlap. A kinetic arrest diagram is constructed, and its
qualitative features agree with the dynamic crossover (MCT) analog. The isothermal
dynamic fragility varies over a wide range, and soft particles are predicted to behave as
strong glasses. The highly variable dependences of the relaxation time on temperature
and volume fraction are approximately collapsed onto two distinct master curves.
The present work serves as a starting point for the development of a predictive
theory of multiple phenomena in real microgel suspensions such as the linear variation of
the shear modulus with volume fraction above RCP, distinctive nonlinear rheology, and
dynamic heterogeneity effects where real world complications, such as osmotic
deswelling and faceting, are important. The present work also motivates further
79
generalization of NLE theory to treat block copolymer micelles and the role of attractive
forces for soft colloids.
We have applied NLE theory to study how particle softness influences the elastic
shear modulus, the connections between the modulus (a short time property) and
activated relaxation (a long time property), and the nonlinear rheological effects of stress-
induced yielding, shear thinning of the relaxation time and viscosity, and stress versus
shear rate flow curves of the repulsive Hertzian contact model of soft sphere fluids.
Below the soft jamming threshold, the shear modulus roughly follows a power law
dependence on volume fraction over a narrow interval for stiff enough particles with an
apparent exponent that grows with repulsion strength. The shear modulus varies inversely
with the transient localization length under all conditions studied. For the barrier and
alpha relaxation time, to a first approximation these local properties are controlled by a
single coupling constant determined by local fluid structure which quantifies the effective
mean square force on a tagged particle.
In contrast to the behavior of hard spheres, the NLE theory for Hertzian spheres
predicts an approximately linear relation between the elastic modulus and activation
barrier. This suggests a microscopic foundation for the often observed, in both thermal
liquids [5] and soft microgel suspensions [17], connections between a (relatively) short
time/distance property and the long time/length scale relaxation process. Moreover, the
predicted proportionality is qualitatively of the elastic shoving model form [5]. A
consequence of this connection is a linear relation between the kinetic volume fraction
and shear modulus, consistent with a recent experiment on microgel suspensions [6].
80
Yielding, shear and stress thinning of the alpha relaxation time and viscosity, and
flow curves were studied as a function of volume fraction and particle stiffness. Yield
strains were found to be relatively weakly dependent on volume fraction and single
particle stiffness for parameters relevant to typical microgel suspensions. Shear thinning
is predicted to commence at Peclet numbers far less than unity, a signature of relaxation
via stress-assisted activated barrier hopping. Power law thinning of the viscosity over
many orders of magnitude of shear rate is generically predicted but with an apparent
exponent smaller than unity. Approaching the soft jamming threshold, a nearly universal
master flow curve can be constructed at fixed repulsion strength which exhibits a power
law form over many intermediate orders of magnitude of reduced shear rate. The breadth
and apparent exponent in this regime systematically increases and decreases, respectively,
as repulsion strength and/or volume fraction grows. Some of the theoretical results are
consistent with microgel experiments, some are not. To achieve a better understanding in
the absence of real world complications, we suggest new computer simulations be
performed for the repulsive Hertzian spheres under strong flow conditions.
4.10 References
[1] For excellent recent reviews of microgel suspensions, see: M.Cloitre, in “Microgel-
based Materials”, H.Wyss, A.Fernandez de las Nieves, J.Mattson and D.A.Weitz Eds;
Wiley-VCH, 2010; R.Bonnecaze and M.Cloitre, Adv.Polym.Sci, 236, 117 (2010).
[2] J.R.Seth, M.Cloitre and R.T.Bonnecaze,J.Rheology, 50, 353 (2006); 52, 1241 (2008).
[3] L.Berthier and T.A.Witten, Europhys.Lett., 86, 10001 (2009); Phys.Rev.E, 80,
021502 (2009).
81
[4] C.Likos, Soft Matter, 2, 478 (2006); C.Likos, Phys.Rep., 348, 267 (2001).
[5] J. Dyre, Rev. Mod. Phys., 78, 953 (2006).
[6] E.H.Purnomo, D.van den Ende, S.A.Vanapili and F.Mugele, Phys.Rev.Lett., 101,
23801 (2008).
[7] J.R.Seth, M.Cloitre and R.T.Bonnecaze,J.Rheology, 50, 353 (2006); 52, 1241 (2008).
[8] Z.Zhang, N.Xu, D.Chen, P.Yunker, A.M.Alsayed, K.B.Aptowicz, P.Habdas, A.J.Liu,
S.R.Nagel and A.G.Yodh, Nature, 459, 83 (2009).
[9] B.M.Erwin, M.Cloitre, M.Gauthier and D. Vlassopoulos, Soft Matter, 12, 2825
(2010); B.M.Erwin, D.Vlassopoulos and M.Cloitre, J.Rheology, 54, 915 (2010).
[10] J.P.Hansen and I.R. McDonald, “Theory of Simple Liquids” (Academic Press,
London, 1986).
[11] L.Berthier, E.Flenner, H.Jacquin and G.Szamel, Phys.Rev.E, 81, 031505 (2010).
[12] H.Jacquin and L.Berthier, Soft Matter, 6, 2970 (2010).
[13] L.Berthier, A.J.Moreno and G.Szamel, Phys.Rev.E, 82, 060501 (2010).
[14] K.S.Schweizer and G.Yatsenko, J.Chem. Phys., 127, 164505 (2007).
[15] K.S.Schweizer, J.Chem.Phys.,127, 16506 (2007).
[16] J.Yang and K.S.Schweizer, Europhys.Lett., 90, 66001 (2010).
[17] J.Mattson, H.M.Wyss, A.Fernando-Nieves, K.Miyazaki, Z.Hu, D.R.Reichman and
D.A.Weitz, Nature, 462, 83 (2009).
[18] K.S.Schweizer and E.J.Saltzman, J.Chem.Phys., 119, 1181(2003).
[19] K.S.Schweizer, Curr.Opin.Coll.Inter.Sci., 12, 297(2007).
[20] G.Yatsenko and K.S.Schweizer, Phys. Rev. E, 76, 041506 (2007); R.Zhang and
K.S.Schweizer, Phys. Rev. E, 80, 011502 (2009).
82
[21] M.Tripathy and K.S.Schweizer, Phys.Rev. E, 83, 041406 and 041407 (2011).
[22] W.M.van Megen and S.Underwood, Phys.Rev.E, 47, 249 (1993).
[23] G.Brambilla, D.ElMasri, M.Pierno, L.Berthier, L.Cipelletti, G.Petekidis and
A.B.Schofield, Phys.Rev.Lett., 102, 085703 (2009).
[24] S.K.Kumar, G.Szamel, and J.F.Douglas, J.Chem.Phys.,124,214501(2006);
Y.Brumer and D.R.Reichman, Phys.Rev.E, 69, 041202 (2004).
[25] V.Kobelev and K.S.Schweizer, Phys. Rev. E, 71, 021401 (2005).
[26] R.C.Kramb, R.Zhang, K.S.Schweizer and C.F.Zukoski, Phys. Rev. Lett., 105,
055702 (2010); J.Chem. Phys., 133, 104902 (2011).
[27] K.Chen, E.J.Saltzman and K.S.Schweizer, J. Phys.-Condensed Matter, 21, 503101
(2009); Annual Reviews of Condensed Matter Physics, 1, 277 (2010).
[28] M.Cloitre, Macromol.Symp., 229, 99 (2005).
[29] M.Cloitre, R.Borrega, F.Monti and L.Leibler, C.R.Physique, 4,221 (2003).
[30] M.Rubinstein and R.H.Colby, Polymer Physics (Oxford Press, Oxford, 2003).
[31] A.LeGrand and G.Petekidis, Rheol.Acta, 47, 579 (2007).
[32] J.C. Dyre and W. H. Wang, J.Chem. Phys., 136, 224108 (2012).
[33] K.S.Schweizer and G.Yatsenko, J.Chem. Phys., 127, 164505 (2007); K.S.Schweizer,
J.Chem.Phys.,127, 16506 (2007).
[34] V.Carrier and G.Petekidis, J.Rheology, 53, 245 (2009).
[35] M.Cloitre, R.Borrega, F.Monti and L.Leibler, Phys.Rev.Lett., 90,068303 (2003).
[36] S.P.Meeker, R.T.Bonnecaze and M.Cloitre, Phys.Rev.Lett., 92, 198302 (2004).
[37] H.Eyring, J.Chem.Phys. 4, 283 (1936).
[38] V.Kobelev and K.S.Schweizer, Phys. Rev. E, 71, 021401 (2005).
83
[39] V.Kobelev and K.S.Schweizer, J.Chem. Phys., 123, 164902 and 164903 (2005).
[40] E.J.Saltzman, G.Yatsenko and K.S.Schweizer, J.Phys.-Condensed Matter, 20,
244129 (2008).
[41] R.Besseling, E.R.Weeks, A.B.Schofeld and W.C.K.Poon, Phys.Rev.Lett. 99, 028301
(2007).
84
4.11 Figures
Fig. 4. 1
Radial distribution functions for 11718E = at volume fractions (bottom to top): 0.3, 0.5,
0.7, 0.9 and 1.1. Inset: Structure factor for 11718E = at volume fractions 0.3, 0.5, 0.7,
0.9 and 1.1 (first peak position from left to right).
85
Fig. 4. 2
Cage peak amplitude, g1, versus differential volume fraction C
φ φ− for (bottom to top):
E = 100 (squares), 1000 (circles), 5832 (up triangles), 11718 (down triangles),
510 (diamonds), 610 (left triangles), 710 (stars). Inset: Zero wavevector value of the
structure factor plotted as 0.75
0S versus volume fraction; the linearly extrapolated value
indicates the RCP state based on the HNC closure.
86
Fig. 4. 3
NMCT ideal glass transition boundary (solid squares) and soft jamming boundary (solid
star) in the format of inverse dimensionless repulsion strength (reduced temperature)
versus volume fraction. The curve through the NMCT points is a parabolic critical power
law fit. Inset: Analogous NMCT ideal glass transition boundary extended to very high
volume fractions where re-entrant behavior is predicted.
87
Fig. 4. 4
Dynamic free energy as a function of particle displacement for E=11718 at volume
fractions (from top to bottom) just below the NMCT crossover and when the barrier is ~5
kBT and 10 kBT. The hard sphere result is the dashed (black) curve for a 10 kBT barrier.
88
Fig. 4. 5
Localization lengths (filled symbols) and barrier locations (open symbols) as a function
of volume fraction for three repulsion strengths: E=4259 (black squares), 5832 (red
circles) and 11718 (green up triangles). Hard sphere results are also shown: localization
length (solid black) and barrier location (dashed red).
89
Fig. 4. 6
Dynamic free energy barrier height (in units of the thermal energy) as a function of
volume fraction for various repulsion strengths (from top to bottom): hard sphere,
E=30000, 20000, 10000, 8000, 5000, and 3000.
90
Fig. 4. 7
Dynamic barrier as a function of the dimensionless coupling constant for three values of
E of 4259 (red, open circles), 5832 (green, open triangles) and 11718 (blue, open
diamonds). The hard sphere result (black, filled squares) is also shown.
91
Fig. 4. 8
Dynamical vertex (Eq.(2.9)) at a fixed barrier height of 10 kBT for (from bottom to top)
the hard sphere and two soft spheres (E=11718 and 5832).
92
Fig. 4. 9
Dimensionless mean barrier hopping (relaxation) time as a function of volume fraction
for repulsion strengths (from top to bottom): hard sphere, E=30000, 20000, 10000, 8000,
5000, and 3000.
93
Fig. 4. 10
Dimensionless mean relaxation time versus repulsion strength for (from bottom to top at
high E): φ =0.55, 0.6, 0.65, 0.7, 0.75, 0.8, 0.85, 0.9.
94
Fig. 4. 11
Determination of the exponent of the effective hard sphere mapping in Eq(10) for soft
repulsive particles based on two vitrification criteria: 2/ 10hop s
τ τ = (black squares) and
4/ 10hop s
τ τ = (red circles).
95
Fig. 4. 12
Collapse onto two master curves of the theoretical relaxation time calculations of
Fig.4.11 based on Eq.(4.5). The dashed line has unit slope.
96
Fig. 4. 13
NMCT ideal glass transition boundary (solid squares), and kinetic arrest curves based on
vitrification criteria of 2/ 10hop s
τ τ = (open square), 310 (open triangle), 510 (open
diamond), in the format of dimensionless temperature versus volume fraction. Curves
through the NMCT and kinetic vitrification points are parabolic critical power law fits.
97
Fig. 4. 14
Fragility plot for a kinetic vitrification criterion of 3/ 10hop s
τ τ = and (from bottom to top)
repulsion strength (E) values of: infinity (hard sphere), 30000, 16000, 8000, 6000. Inset:
Fragility as a function of scaled temperature for the three indicated kinetic arrest criteria.
98
Fig. 4. 15
Log-log plot of the dimensionless shear modulus (units of kBT/σ3) as a function of
volume fraction for various Hertzian potential dimensionless strengths of (from top to
bottom): hard sphere, E=30000, 20000, 10000, 8000, 5000, 3000, 2000, 1000, 800.
99
Fig. 4. 16
Dimensionless shear modulus divided by volume fraction versus the coupling constant
( 2
1gφ , where g1 is the cage peak amplitude of the pair correlation function) for E = 4259
(red, open circles), 5832 (green, open up triangles) and 11718 (blue, open diamonds).
Hard sphere results are shown as solid squares. Inset: * 1 2( / )LOC
G rφ σ− as a function of
volume fraction for the same three soft particles examined in the main frame.
100
Fig. 4. 17
Log-log plot of the dynamic free energy barrier (units of thermal energy) as a function of
dimensionless shear modulus divided by volume fraction for the same three soft particle
systems as in Figure 2 (E = 4259 (red, dashed lines), 5832 (green, dotted lines) and 11718
(blue, dash-dot lines). Hard sphere results are also shown as black, solid lines. Inset:
analogous linear-linear plot.
101
Fig. 4. 18
Kinetic glass transition volume fraction, g
φ , versus the dimensionless shear modulus
based on the hopping time vitrification criterion 3/ 10hop s
τ τ = . Inset: g
φ as a function of
the corresponding single particle dimensionless repulsion strength, E.
102
Fig. 4. 19 (A)
Barrier height divided by its zero stress value as a function of applied stress
nondimensionalized by the absolute yield stress for three E values and two volume
fractions (solid, 0.7; dashed, 0.9). The hard sphere result is also shown.
103
Fig. 4. 19 (B)
Linear plot of the dimensionless absolute yield stress as a function of volume fraction
(From top to bottom: hard sphere, E=11718, E=5832, E=4259). The inset shows the same
results in a log-log format.
104
Fig. 4. 20
Two measures of yield strain: absolute (solid curves) and mixed (dashed curves) as a
function of volume fraction for E =4259, 5832, 11718, from top to bottom at large φ.
105
Fig. 4. 21
Mean barrier hopping time normalized by its zero stress quiescent value as a function of
dimensionless applied stress for E=4259, 5832, 11718 (from top to bottom) and two
volume fractions (solid, 0.7; dash, 0.9). Inset: Same results but the dimensionless hopping
time is expressed in units of the short relaxation time for E=4259, 5832, 11718 (from
bottom to top) and two volume fractions (solid, 0.7; dash, 0.9).
106
Fig. 4. 22
Dimensionless shear stress as a function of shear rate nondimensionalized by the short
relaxation time for various volume fractions: 0.6 (solid), 0.7 (dash), 0.8 (dot), 0.9 (dash-
dot), and E=4259 (black square), 5832 (red circle), 11718 (green triangle).
107
Fig. 4. 23
Flow curve in the representation of stress reduced by the corresponding absolute yield
value versus shear rate reduced by a time scale involving the short time relaxation
process and the dimensionless glassy shear modulus. Results are shown for a single
E=11718 at six volume fractions. Note the near collapse in the high volume fraction soft
jammed regime (J
φ φ> ), and the lack of collapse below this crossover volume fraction.
108
Fig. 4. 24
Shear viscosity reduced by its quiescent value as a function of shear rate reduced by the
quiescent relaxation time (i.e., Peclet number). Results are shown for E=11718 and six
volume fractions.
109
CHAPTER 5
THEORY OF THE STRUCTURE AND MISCIBILITY
OF SOFT FILLER POLYMER NANOCOMPOSITES:
METHODOLOGY
5.1 Introduction
The addition of nanoparticles or fillers to dense polymer melts can profoundly
modify the mechanical, thermal, optical and/or other material properties of the resulting
polymer nanocomposite (PNCS) because the large surface to volume ratio of those
nanoparticles. Nanoparticles can be of very different shapes, the simplest and most
widely studied are spheres. However, in order to effectively tune the materials properties
often requires nanoparticle dispersion, which is a challenging topic of intensive studies.
[1] The major physical reasons for the aggregation of nanoparticles lies on the facts that
the large entropy penalty to have polymers packing around nanoparticle, which generate
strong depletion attractions among nanoparticles [2-5], and direct Van der Waals
attractions if the dielectric constants are mismatched. There are multiple potential ways to
avoid particle aggregation, such as adding surfactant or polymer brushes to add
interfacial repulsions between nanoparticles to counteract depletion attraction, or
engineer polymer-particle attraction of appropriate strength. Recently, the consequence of
surface morphology of spherical fillers has been studied experimentally [6]. It is possible
to disperse soft spherical nanoparticles, which have a rough, fluctuating surface, in
chemically matched homopolymer melt, which is not true if the filler is a smooth hard
sphere with no polymer-particle attraction [7]. A generic key fundamental issue here is
110
how the nanoparticle surface corrugation and fluctuation modify the statistical spatial
organization of the nanoparticles in a dense polymer matrix.
From a theoretical point of view, computer simulations of nanoparticles dissolved
in a dense polymer melt are an attractive option. However, they are computationally very
expensive and difficult to equilibrate due to the relevant high total packing fraction, large
size asymmetry between monomers and fillers, and chain connectivity constraints.
Significant progress has been made in the development of approximate statistical
mechanical theories for polymer nanocomposites during the last decade based on the
microscopic integral equation approach known as the Polymer Reference Interaction Site
Model (PRISM) theory [8], which accounts for packing effects on the monomer and
beyond scale, and the strength and range of material-specific attractive and repulsive
interactions. PRISM theory has been extensively applied to the problem of spherical
fillers in polymer melts and dense solutions [7,9,10].
Our present work is partially inspired by recent experiments on the dispersion of
crosslinked polystyrene nanogels in chemically matched dense polymer matrix of linear
chain-like polystyrene (monomer size~1 nm) [6]. These nanoparticles are intramolecular
crosslinked polymeric networks with diameters that range from ~5 nm to 7 nm or larger
depending on the pre-polymer molecular weight (before crosslinking process) and the
degree of intramolecular crosslinking [6]. The Kratky plots obtained from neutron
scattering data in solution show a shift toward particle-like nature for heavily crosslinked
nanogels (20 mol % crosslinker) [11]. Based on viewing the nanogel as a hard sphere, it
has been speculated that this entropically unfavorable dispersion is offset by an enthalpy
gain due to an increase in molecular contacts at dispersed nanoparticle surfaces as
111
compared with the surfaces of phase-separated nanoparticles [6]. No direct evidence of
any kind exists for this explanation. The nanoparticles are largely, but not entirely,
impenetrable, and have been widely studied due to their importance as advanced
functional materials, e.g. the nanoparticles alone can form liquids and gels under various
conditions and undergo surface segregation in thin films [12]. Single nanogel softness or
elasticity is highly variable via manipulation of solvent quality and crosslink density.
This feature makes the nanogels distinctive from its counterpart like rough and rigid hard
sphere fractals [13]. By definition, the surface of such soft nanoparticles must be fuzzy
and fluctuating. These facts serve as our major motivation, but we can investigate much
broader classes of surface rough nanoparticle schemes, e.g. rigid but rough carbon black
nanoparticles [13], nanoparticle coated liposome (Pickering emulsion) [14], and repulsive
polymer-tethered colloids [15], using the following methodology.
Our goal is to develop a new hybrid small scale Monte Carlo (MC) simulation
plus Polymer Reference Interaction Site Model (PRISM) theory. We first construct a
minimalist multi-scale model, which explicitly studies the role of surface corrugation and
fluctuation by introducing the dimensionless length scales for particle-monomer size
asymmetry, surface corrugation and fluctuation magnitude. Then we perform small-scale
MC simulation to correctly capture the effective interactions among particles and
monomers. We then employ PRISM theory to study the spatial statistical correlation
functions of the mixture to understand how the surface morphology affects depletion
attraction and bridging and whether it will stabilize dispersion. The following studies are
done in two major categories: (1) the athermal limit, i.e. all local interactions are hard-
112
core interactions, and (2) when interfacial polymer particle attractions exist. For both
categories, we study systems that have either a frozen surface or a fluctuating surface.
It is important to realize that every surface has corrugation. Even for a silica
particle, it may not be smooth at atomic scale, but it is very likely to be smooth at the
monomer (~nm) length scale. Another example is carbon black widely used in tire
industry. They are rigid and rough nanoparticles crucial to tune the mechanical responses
of tires. It is hard to know at what level of description the modeling of surface roughness
matters most. We do not aim to address every aspect of the problem but to offer one new
computational approach that accounts for effect of the surface corrugation and fluctuation
in a unified way.
The remainder of this chapter is structured as follows. Section 5.2 discusses the
new minimalist multi-scale model of soft nanoparticles and the setup of small-scale
Monte Carlo simulations: bead-core model, the Gaussian distribution of bead position
fluctuation, effective interaction between two nanoparticles in a vacuum, and the
effective interaction between one nanoparticle and one monomer in a vacuum. A review
of PRISM theory is given in section 5.3. The theoretical details of the extension of
PRISM theory to treat soft, interfacially rough filler PNCS and three-step approach are
presented in section 5.4. Section 5.5 and 5.6 present the references and figures,
respectively.
113
5.2 Minimalist Model of Soft Nanoparticles and Effective Interactions based on
Small Scale Monte Carlo Simulation
Our goal is not to describe exactly the morphology of any particular real soft
nanoparticles in a particular polymer melt or solvent. Rather, we aim to construct
representative but minimalist models that encode the most basic aspects of surface
corrugation and fluctuation and extreme limiting cases of dynamic surface fluctuation.
Crosslinked nanogels of size ranges from 5 nm to 10 nm serve as motivation for model
construction. These soft particles have many features, including characteristic length
scales like surface corrugation size and fluctuation magnitude, related to well controlled
crosslink densities.
5.2.1 Bead-core model of spherical nanoparticle with static corrugation
We characterize the nanoparticle surface roughness by constructing a core-bead
model. The particle has a hard core of diameter D, and there are N spherical beads
centered at the surface of the core and covering the surface (see Fig.5.1). The bead
diameter isσ . The relative ratio of core size and bead size, /D σ , characterizes the static
corrugation. For geometry reasons, there are only a limited set of options of ratio to
choose so that the surface is densely covered by the beads [16]. In our studies, unless
stated otherwise, we use 72 because this number of beads is needed to achieve / 5D σ = ,
which gives a ratio relevant for real world nanogels (particle size~ 5nm and monomer
size ~1 nm). Other cases studied are 162 and 282, which gives relative ratio of
/ 7.8D σ = and / 10D σ = , respectively.
Instead of a full site level description of the soft filler that includes both the
surface corrugation length scale as well as all the internal degree of length scales, here we
114
use a hard core to model all internal buried site level information. As a result, we assume
all the internal sites avoid particle or monomer penetration beyond the surface
corrugation level. This approximation can help us to avoid conceptual and computation
problems while keeping the most important feature for describing depletion phenomena
that is largely dependent on length scales of order one half a monomer diameter.
5.2.2 Gaussian distribution of bead position fluctuation
We introduce surface fluctuation by allowing the beads to fluctuate radially (no
transverse motion allowed) and incoherently (beads positions are fluctuating randomly
and not in the same phase), obeying a Gaussian distribution function:
2
2
( )
( , ) ,0
r R
up r u e r
−−
∝ < < ∞ (5.1)
where p(r,u) is the probability to find a bead at distance r from the origin of the core of
radius R=0.5D. The parameter u quantifies the magnitude of radial position fluctuation,
which has a dimension of length and is related to a statistically sensible mean squared
fluctuation magnitude, which we take to be small and less than / 2σ . The latter length is
of order the range of the polymer mediated entropic depletion attraction. Moreover, we
desire to maintain particle shape stability. The assumption of incoherent fluctuation
generates maximal dynamic disorder in surface roughness. The relation between the
statistical mean squared fluctuation magnitude and u is:
2 2
0
( ) ( ) ( , )r R r R p r u dr
∞
< − > = −∫ (5.2)
Some representative numbers used in our calculation are shown below:
115
/u σ 2( ) /r R σ< − >
0 0
0.0625 0.04
0.125 0.087
0.25 0.17
We also study another limiting model: the beads can fluctuate in a coherent
manner (polar opposite), which we believe is less realistic compared with the incoherent
fluctuation model. Unless otherwise indicated in our following discussions, the
incoherent fluctuation model is employed.
5.2.3 Small-scale Monte Carlo simulations and effective interactions
The center-of-mass (CM) level effective interaction description within an
effective one-component fluid model has been extensively utilized for ‘‘soft colloids’’
such as many arm stars, crosslinked microgels and block copolymer micelles [17-21], and
also charged colloidal stars, branched polyelectrolytes and carbon black fractals [22,23].
CM-level coarse-grained models of water have also been constructed and studied [24,25].
The idea is to average over the internal degrees of freedom at the two “particle” level to
obtain a CM effective interaction. The nanoparticles of present interest are interfacially
rough, spherical particles, with/without surface fluctuation/interfacial cohesion. These
features, in conjunction with surface roughness, result in an effective interaction that
displays soft repulsions or attractions. Our current research makes one-step forward
compared to typical one-component fluid models by utilizing these CM level interactions
in an effective two-component mixture calculation.
We now discuss details of the dilute two-particle effective potential calculation at
the center-of-mass level, ( )eff
nnU r . The cross potential ( )eff
mnU r , of the dilute one particle
one monomer effective potential calculation is similarly treated. The nanoparticles can
116
rotate and adopt different orientations characterized by two angles ,θ φ . The conditional
configurational partition function for two particles at fixed CM separation, r, is written as
an integral over the two angles of rotations of each aggregate, ( ,θ φ ), where, in general, it
is a function of these angles:
1 2 1 2( , , , , )
1 2 1 1 2 22
1( ) sin sin
4
E rZ r d d d d e
β θ θ φ φθ θ θ φ θ φπ
−= ∫ (5.3)
is computed using standard multi-dimensional integration methods akin to an elementary
Monte Carlo integration:
1, 2, 1, 2,
1, 2, 1, 2,
2( , , , , )
, ,min , ,min 1, 2,211
2( , , , , )
1, 2,1
1 1( ) ( )( ) sin sin
(4 )
1( ) sin sin
4
n n n n
n n n n
NE r
i max i i max i n nni
NE r
n nn
Z r eN
Z r eN
β θ θ φ φ
β θ θ φ φ
θ θ φ φ θ θπ
πθ θ
−
==
−
=
= − − ∑∏
= ∑
i
(5.4)
where ( 1, 2, 1, 2,, , ,n n n n
θ θ φ φ ) are four uniform random numbers falls in the range within
integration limits. Consider the athermal limit as an example, which means that the all
local interactions are hard-core at the elementary site level. To calculate the effective
potentials for particles with fluctuating surfaces, we need to average out the fluctuation
degrees of freedom in addition to the above rotational degrees of freedom.
1 2 1 2
2( , , , , )
1 21
1( ) sin sin
4
NE r
n
Z r eN
β θ θ φ φπθ θ −
=
=< >∑ (5.5)
where the average is done with respect to the bead position probability distribution
function 1 72 1 72
1 1 1 2 2 2( ,... ,... , ,... ,... , )i jp r r r r r r u . This is computed using standard multi-
dimensional integration methods akin to an elementary Monte Carlo integration:
1 2 1 2
2( , , , , , )
1 21 1
1 1( ) [ sin sin ]
4
i
p
M NE r r
m n
Z r eM N
β θ θ φ φπθ θ
−
= =
≅ ∑ ∑ (5.6)
117
where 1 72 1 72
1 1 1 2 2 2 ,... ,... , ,... ,... i i i
pr r r r r r r= are independent Gaussian random numbers
indicating the distance between the bead i (or j) on particle 1(or 2) and its core origin.
The probability to find two particles separated by a distance in the interval r and r+dr
is( )2 2( ) 4 4 ( )
eff
nnU r
P r dr r dre r drZ rβπ π−∝ ∝ , where ( )eff
nnU r is the desired effective
potential. By exploiting spherical symmetry, one can then write the potential as:
1 2 1 2
2( , , , , , )
1 21 1
2
1 2
1 1( ) ln[ ( )] ln [ sin sin ]
4
1 1( ) ln [ sin sin ]
4
i
p
m n
M NE r reff
nn B Bm n
eff
nn Bnoc noc
U r k T Z r k T eM N
U r k TM N
β θ θ φ φπθ θ
πθ θ
−
= =
= − ≅ − ∑ ∑
= − ∑ ∑
(5.7)
where ‘‘noc” stands for ‘‘non-overlapping configurations’’. We use the following rule to
define overlapping configurations: any part of particle 1 (core or bead) overlaps with any
part of particle 2 (core or bead). Here the summation is equivalent to counting noc
because:
0,
,
E noc
E otherwise
=
= +∞ (5.8)
To calculate ( )eff
mnU r , one needs two uniform random numbers ( ,θ φ ) and 72
independent Gaussian random numbers, and the same procedure as described above is
followed. An example of the number of configurations we used to get accurate results is
M=1000, N=1000. Therefore, the total number of configuration is 610 .
5.2.4 Interfacial attraction at surface corrugation level ( )eff
mnU r
The chemical nature of the polymer and nanoparticle is encoded partially in size
asymmetry (D/d and/or /d σ if the particle surface is rough) and partially in effective
pair potentials: ( )eff
mmU r , ( )eff
nnU r , ( )eff
mnU r , where m stands for monomer and n stands for
(CM) nanoparticle. If all these interactions are hard core, then the polymer
118
nanocomposite is effectively an athermal system. However, a more common situation is
when ( )eff
nnU r and ( )eff
mmU r are effectively hard core interactions and ( )eff
mnU r is composed
of a repulsive branch (often hard core) and longer-range effective attractive interactions:
,
exp[ ],
eff
mn mn
eff mn
mn mn
U r
rU r
σ
σε σ
α
= ∞ <
−= − − >
(5.9)
where ε is the contact cohesion strength, α is the attraction range and ( ) / 2mn
D dσ = + .
Based on prior PRISM theory studies of such a nanocomposite with smooth hard sphere
fillers [26], in this enthalpic mixture where 1βε >> , strong enough attractive
interactions between the filler and monomers can result in thin layers of polymer strongly
associating with, or adsorbing onto, the particles. This results in a particle-particle
potential of mean force which favors well-defined, small interparticle separations, i.e.,
local “bridging”. For 1βε << , the entropic depletion attraction is dominant and
nanoparticles contact aggregate, leading to classic macrophase separation. Between these
two extremes is a third behavior, in which a polymer gains enough cohesive interaction
energy to associate with a single filler, but not enough to give up the additional entropy
required for association with multiple particles. In this case, a nanoparticle is surrounded
by a thermodynamically stable “adsorbed” polymer layer, typically on the order a few
monomer diameters thick, which sterically stabilizes the particles in the polymer matrix.
For all situations, polymer degree of polymerization is a second order effect due to the
high total packing fraction of a dense PNCs.
In our multi-scale modeling where particles have two length scales (D, σ ) instead
of one, in ( )eff
mnU r the nature of the interfacial attraction is related with the surface
119
corrugation and bead size. This is more realistic in the sense that the interfacial cohesion
has to be related to the interface which is mimicked by corrugated beads. As a result, we
define the bare interaction at bead-monomer level as:
,
exp[ ],
mb mb
mb
mb mb mb
mb
E r
rE r
σ
σε σ
α
= ∞ <
−= − − >
(5.10)
where mb
ε and mb
α are the attraction strength and attraction range between the polymer
monomer and surface bead, and ( ) / 2mb
dσ σ= + . Thus, the effective interaction between
one nanoparticle and one monomer is modified as: for frozen surface,
2( , , )
1
1( ) ln[ sin ]
4mb
NE reff
mn Bn
U r k T eN
β θ φπθ −
=
≅ − ∑ (5.11)
for fluctuating surface,
1
2( , , , )
1 1
1 1( ) ln [ sin ]
4
i
mb
M NE r reff
mn Bm n
U r k T eM N
β θ φπθ −
= =
≅ − ∑ ∑ (5.12)
5.3 Equilibrium Theory of Polymer Nanocomposites
The Polymer Reference Interaction Site Model (PRISM) integral equation
approach [8] has been recently extended to treat the structure, thermodynamics and
phase behavior of mixtures of hard particles and homopolymers in solutions [9] and
nanocomposite melts [10]. Objects of arbitrary shapes are represented as bonded sites
that interact via pair decomposable site–site potentials. If a species is rigid, then
computation of the corresponding intramolecular pair correlation function is a simple
exercise in geometry. For flexible polymers, the intra-chain structure is described
statistically. It is either approximated by an ideal (in the global Flory sense) coil at
120
various levels of chemical realism (e.g., Gaussian, freely-jointed, semiflexible, or
rotational isomeric chain), or determined in a fully self-consistent manner with
intermolecular packing correlations based on a medium-induced solvation potential and
single chain Monte Carlo simulation [8,27]. The chemical nature of the polymer and
nanoparticle is encoded in pair potentials, ( )ij
U r , composed of a repulsive branch
(often hard core) and longer-range attractive interactions (if there is interfacial
cohesion). The theory is defined by a set of coupled nonlinear matrix integral equations
, ,, , ,, , ,, ,, , ,, , ,, , ,, ,,( ) (| |) (| |) ( ) (| |) (| |) ( )H r dr dr r r C r r r dr dr r r C r r H r= Ω − − Ω + Ω − −∫ ∫ ∫ ∫
(5.13)
Here, ( )H r
contains elements ( ( ) 1)i j ij
g rρ ρ − , j
ρ is the corresponding site number
density, ( )ij
g r is the intermolecular site-site radial distribution function, which describes
how the density varies as a function of the distance r from a reference site. ( )C r
contains
elements ( )ij
C r , which are renormalized site–site intermolecular potentials or “direct
correlation” functions. ( )rΩ
contains elements ( )ij
rω , which defines the intramolecular
pair correlations (species statistical shape) and is linear in site number densities. The
intramolecular correlation function obeys:
1
1, 1
1
1, 1
( ) ( | |),
( ) ( ) ( | |),
N
iji j
N
iji j
r N r r
r N N r r
α
α
αγ α
αγ α γ
ω δ α γ
ω δ α γ
−
= =
−
= =
= − =∑
= + − ≠∑
(5.14)
where ,α γ correspond to different components, i.e. monomer or nanoparticle. The first
term on the right hand side of equation. (5.13) describes all intermolecular correlations in
the dilute limit, and the second term quantifies many body effects.
121
To render the number of integral equations tractable, polymer chain end effects
are pre-averaged [8]. For example, for a mixture of hard spheres (A) and single site (B)
homopolymers, there are three intermolecular pair correlation functions: ( )AA
g r , ( )AB
g r ,
and ( )BB
g r . Even if the intramolecular correlations are known, one requires an
approximate closure which provides a second set of relations between the pair and direct
correlation functions and bare potentials. Reliable closures for polymer–particle systems
are not obvious due to the large structural and packing asymmetry between flexible
chains and hard particles. The classic site–site Percus–Yevick (PY) closure [28] is
employed for polymer–polymer (p–p) and polymer–nanoparticle (p–n) correlations,
( )( ) (1 ) ( )ij
U r
ij ijC r e g r
β= − (5.15)
and the hypernetted chain (HNC) closure is adopted for nanoparticle–nanoparticle (n– n)
correlations:
( ) ( ) ( ) ln ( )nn nn nn nn
C r U r h r g rβ= − + − (5.16)
Numerical solution of above equations yields ( )ij
g r , and collective concentration-
concentration fluctuation partial scattering structure factors in Fourier space, ( )ij
S k ,
describing correlated concentration fluctuations over all length scales. Of special
interest is the dilute two particle potential of mean force (PMF):
( ) ln( ( ))nn nn
W r g rβ = − (5.17)
from which second virial coefficients can be deduced. The number of coupled equations
increases as number of components increases, which results in great numerical challenges
to solve these coupled nonlinear integral equations. [8] We note that PRISM theory is not
a mean field theory since it predicts correlations over all lengths scales (which are
122
coupled) within a compressible fluid framework. The intermolecular structure is
functionally related to the intramolecular structure and bare interparticle interactions.
However, PRISM theory only treats spatially homogeneous states. For dense PNCs, most
PRISM studies to date have employed a freely-jointed chain (FJC) homopolymer model
(degree of polymerization, N) composed of sites of diameter d, and hard spherical
particles (diameter D), under fixed melt-like total packing fraction conditions.
The packing fraction or volume fraction of a particular species is defined as:
3 , ,6
i i iD i m n
πφ ρ= = (5.18)
where i
D is the size for species i. The total packing fraction or volume fraction:
, ,i
i
i m nη φ= =∑ (5.19)
5.4 Combining Small Scale Monte Carlo Simulation with PRISM Theory: Three-
Step Approach
5.4.1 PNCs: Multi-scale modeling and model parameters setup
Fig.5.2 shows a representation of the proposed multi-scale modeling approach for
soft, interfacially rough filler PNCs. Polymers are treated as athermal chains of N
spherical interaction sites, or monomers, of diameter d that interact via pair-
decomposable hard-core potentials. A freely jointed chain (FJC) model is adopted with a
rigid bond length 4 / 3l d= (corresponding to a persistence length of 4/3), and d is the
polymer segment i.e. monomer diameter. The FJC structure factor is:
2 1 1 1 2( ) [1 2 2 ] /(1 )N
pk f N f N f fω − − += − − + − (5.20)
123
where f =sin(kl)/kl. The FJC chain model ignores nonideal conformational effects, which
are expected to be minor for the melt conditions of interest [29,30]. In principle, filler
perturbation of polymer conformation at nonzero volume fractions could be treated based
on the fully self-consistent version of PRISM theory, which involves the construction of a
medium-induced solvation potential and solution of an effective single-chain problem
with Monte Carlo simulation. [8,27] The accuracy of this approach for polymer
nanocomposites, particularly when the monomers are strongly attracted to the fillers, is
not fully known. However, the few existing experiments and simulations for dense
polymer nanocomposites at finite filler loadings suggest that conformational
perturbations are small or negligible. [29,30] Most importantly for the present work is
that, because our focus is the dilute filler limit, within the liquid-state theory approach the
statistical conformations are not perturbed. We consider the dilute particle limit ( 0n
φ → )
and polymer melt condition ( 0.4η = ).
Nanoparticle fillers are treated as spheres with the surface corrugation beads as
discussed in section 5.2. A. We study core bead size ratios of /D σ = 5 (72 beads), 7.8
(162 beads), and 10 (282 beads).
5.4.2 PRISM theory sequentially solving strategy and closures
Using the two effective potentials obtained from the small scale MC simulations,
we erase all beads from the nanoparticles and perform an effective two-component
PRISM calculation in the dilute particle limit. There, the 3 coupled integral equations,
written here in Fourier space, can be sequentially solved:
2
( ) ( ) ( ) ( )
( ) ( ) ( )
( ) ( ) ( ) ( ) ( ) ( )
pp p pp pp
pn pn pp
nn nn p pn pp nn
h k k C k S k
h k C k S k
h k C k C k S k C k W k
ω
ρ
=
=
= + ≡ +
(5.21)
124
By using PY closure for monomer-monomer and particle-monomer, we can solve the
first two equations and then we perform a Fourier Transform of the last equation and get
( ) ( ) 1 ( )nn nn
g r C r W r= + + (5.22)
The closure for the particle-particle correlation is the HNC. However, numerical
solutions based on using the PY closure is more easily obtained and can be related (when
0n
φ → ) to the solution based on HNC through the following procedure: For soft fillers,
*( ) 0,nn
g r r D= < , where *D D≥ ( *
D D> representing the CM separation distance, at
which no configuration exists that avoids overlaps. ). By using PY closure for particle-
particle direct correlation function, one has ( ) *( ) (1 ) ( ),
eff
nnU rPY PY
nn nnC r e g r r D
β= − > . Thus,
( ) *( ) (1 ( )),eff
nnU rPY
nng r e W r r D
β−= + > . Then the HNC closure for particle-particle
correlations are: *( ) ( ) 1 ln ( ),HNC HNC HNC
nn nn nnC r g r g r r D= − − > . Thus,
( ) ( ) *( ) ,eff
nnU rHNC W r
nng r e e r D
β−= > . We also have
( ) *( ) exp[ ( ) ( ) 1],eff
nnU rHNC eff PY
nn nn nng r U r e g r r D
ββ= − + − > (5.23)
Based on the equation above, one can also write the polymer mediated potential of mean
force as
( )( ) ln ( ) ( ) [ ( ) 1] ( ) ( )
eff
nnU rpolymer HNC eff PY eff
nn B nn nn B nn nn BW r k T g r U r k T e g r U r k TW r
β= − = + − = − (5.24)
5.4.3 Three-step approach
Fig.5.3 presents the conceptual idea of the three-step approach: On step one and
two, we carry out small scale MC simulations to construct the effective pair interactions.
Then on step three, all beads from particles are erased and we use effective interactions to
mimic their effects. Monomers are connected into chains and the “Center of Mass” level
125
nanoparticle (dilute) in homopolymer melt mixtures problem is solved via a standard
two-component PRISM theory methods described above.
A natural question might is why we avoid doing calculations directly targeted on
dealing with three components (core, bead, and monomer) at the same time? The answer
lies in two facts. First, dealing with multiple length scales at the same time involves
solving coupled nonlinear integral equations, which is not a trivial computational
problem and generally is very difficult [31]. Considering the fact that our system has
large size asymmetry (i.e. nanoparticle vs. monomer/bead/fluctuation magnitude) and
four length scales, this problem becomes even more difficult. More importantly, the
three-step approach can be considered as a general solution to treat multi-scale systems,
because of the conceptual reduction of dimensions to treat two length scales at the each
time, we are free to extend our studies to much more complicated systems easily, i.e.
changing the fluctuation manner, local interaction, introducing randomness in corrugation
and etc.
5.5 References
[1] N. S. Sariciftci, L. Smilowitz, A. J. Heeger, F. Wudl, Science 258, 1474 (1992). N. S.
Sariciftci, A. J. Heeger, Synth. Met. 70, 1349 (1995). M. E. Mackay et al., Nat. Mater. 2,
762 (2003). B. Ash, L. Schadler, and R. Siegel, Mater. Lett. 55, 83 _2002_. B. J. Ash, R.
W. Siegel, and L. S. Schadler, J. Polym. Sci., Part B: Polym. Phys. 42, 4371 _2004_. A.
Bansal, H. Yang, C. Li, B. C. Benicewicz, S. K. Kumar, and L. S. Schadler, J. Polym.
Sci., Part B: Polym. Phys. 44, 2944 _2006_. G. Tsagaropoulos and A. Eisenberg,
Macromolecules 28, 6067 _1995_.
126
[2]W. B. Russel, D. A. Saville, and W. R. Schowalter, Colloidal Dispersions, Cambridge
University Press, New York, 1989.
[3] W. C. K. Poon, J. Phys.: Condens. Matter 14, 859, (2002).
[4] S. Asakura and F. Oosawa, J. Chem. Phys. 22, 1255, (1954).
[5] S. Asakura and F. Oosawa, J. Polym. Sci. 33, 183, (1958).
[6] M. E.Mackay, A. Tuteja, P.M.Duxbury, C. J.Hawker, B Van Horn, Z. Guan, G Chen,
R. S. Krishnan, Science 311, 1740 (2006)
[7] J.B.Hooper&K.S.Schweizer,Macromolecules, 38, 8858 (2005)
[8] K. S. Schweizer and J. G. Curro, Adv. Chem. Phys. 98, 1, 1997); J. G. Curro, K. S.
Schweizer, G. S. Grest, and K. Kremer, J. Chem. Phys. 91, 1357 ,1989); K. Honnell, J. G.
Curro, and K. S. Schweizer, Macromolecules 23, 3496 (1990).
[9]M.Fuchs, K.S.Schweizer. J Phys Condens Matter 2002;14:R239–69.
[10]L.M.Hall, K.S.Schweizer. J Chem Phys 128, 234901(2008)
[11] A,TUTEJA, M.E.MACKAY, C. J. HAWKER, B VAN HORN, D. L. HO, Journal of
Polymer Science: Part B: Polymer Physics, 2006
[12] Antonietti, M.; Sillescu, H.; Schmidt, M.; Schuch, H.Macromolecules 1988, 21,
736–742. Antonietti, M.; Ehlich, D.; Folsch, K. J.; Sillescu, H.; Schmidt, M.; Lindner, P.
Macromolecules 1989, 22, 2802–2812. Antonietti, M.; Bremser, W.; Schmidt, M.
Macromolecules 1990, 23, 3796–3805. Antonietti, M.; Bremser, W.; Pakula, T. Acta
Polym 1995, 46, 37–44.
[13] Rajarshi Chakrabarti,a Jean-Yves Delannoy,b Marc Coutyb and Kenneth S.
Schweizer, Soft Matter, (2011), 7, 5397
127
[14] Julius W. J. de Folter , Marjolein W. M. van Ruijven and Krassimir P. Velikov, Soft
Matter, 2012,8, 6807-6815
[15] A. Jayaraman and K. S. Schweizer, J. Chem. Phys. 128 164904 (2008)
[16] R. H. Hardin, N. J. A. Sloane and W. D. Smith, Tables of spherical codes with
icosahedral symmetry, published electronically at
http://www.research.att.com/~njas/icosahedral.codes/.
[17] J. Yang and K. S. Schweizer, Europhys. Lett., 2010, 90, 66001.
[18] W. R. Chen, F. Mallamace, C. J. Glinka, E. Fratini and S. H. Chen,
Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip. Top., 2003, 68, 041402.
[19] J. R. Seth, M. Cloitre and R. T. Bonnecaze, J. Rheol., 2006, 50, 353–376.
[20] F. Ozon, G. Petekidis and D. Vlassopoulos, Ind. Eng. Chem. Res., 2006, 45, 6946–
6952.
[21] C. N. Likos, H. Lowen, M. Watzlawek, B. Abber, O. Jucknischke, J. Allgaier and D.
Richter, Phys. Rev. Lett., 1998, 80, 4450.
[22] A. Jusufi, C. N. Likos and H. Lowen, Phys. Rev. Lett., 2002, 88, 018301.
[23] F. Huang, K. Addas, A. Ward, N. T. Flynn, E. Velasco, M. F. Hagan, Z. Dogic and S.
Fraden, Phys. Rev. Lett., 2009, 102, 108302
[24] Z. Yan, S. V. Buldyrev, P. Kumar, N. Giovambattista and H. E. Stanley, Phys. Rev.
E: Stat., Nonlinear, Soft Matter Phys., 2008, 77, 042201.
[25] W. P. Krekelberg, T. Kumar, J. Mittal, J. R. Errington and T. M. Truskett, Phys. Rev.
E: Stat., Nonlinear, Soft Matter Phys., 2009, 79, 031203.
[26] J.B.Hooper&K.S.Schweizer,Macromolecules, 39, 5133 (2005)
[27] Heine DR, Grest GS, Curro JG. Adv Poly. Sci. 2005;173:209–49.
128
[28] Chandler D. Equilibrium theory of polyatomic fluids. In: Montroll EW, Lebowitz JL,
editors. Studies in statistical mechanics, vol. 8. Amsterdam: North Holland; 1982. p. 275–
340.
[29] J. B. Hooper and K. S. Schweizer, Macromolecules 39, 5133 _2006_.
[30] J. B. Hooper and K. S. Schweizer, Macromolecules 40, 6998 _2007_.
[31] There are 6 coupled nonlinear integral equations to slove for a 3-component PRISM
calculation.
129
5.6 Figures
Fig. 5. 1
Schematic representation of a bead-core model of rigid structured nanoparticles (Number
of beads=72) is shown. There are two length scales: the core size D and the surface bead
size σ . If the beads fluctuate, a third length, the amplitude of vibration enters.
130
Fig. 5. 2
Representation of the multi-scale modeling of soft, interfacially rough polymer
nanocomposites. The chain length, N, is fixed to be 100. There are five length scales:
Particle size (D), surface bead (corrugation) size (σ ), monomer size (d), u in unit of
length, related to surface fluctuation magnitude and polymer radius of gyration
/ 6g
R N d= ⋅ . From previous PRISM studies, the radius of gyration is a secondary
effect, so there are three important dimensionless length scale ratios: /D d , /D σ and
/u σ .
131
Fig. 5. 3
Representation of 3-step approach: Step 1 & 2, we conduct small scale MC simulations to
obtain effective interactions; on step 3, we erase beads on the surface and use effective
interactions to mimic their effects and do an effective two-component PRISM calculation.
132
CHAPTER 6
THEORY OF THE STRUCTURE AND MISCIBILITY OF
SOFT FILLER POLYMER NANOCOMPOSITES:
RESULTS AND DISCUSSION
6.1 Introduction
This chapter presents applications of the methodology developed in the previous
chapter. We start with the smooth hard sphere filler, then add surface beads to mimic
static surface corrugation and vary it by changing the number of beads on the surface.
Surface fluctuation is then added by allowing beads to fluctuate radially either in an
incoherent or coherent manner to vary the “dynamic” surface corrugation. Fig.6.1
illustrates the types of systems studied. All these initial studies are for the athermal
mixture, which is the chemically matched mixture of high interest for experimentalists,
e.g. crosslinked polystyrene and linear polystyrene chain mixture [1]. Finally, interfacial
cohesion is added and its consequences are studied. All calculations are for the incoherent
surface fluctuation model unless otherwise noted.
Section 6.2 presents the effective interaction between two nanoparticles, and the
effective interaction between one nanoparticle and one monomer, in a vacuum, for both
conditions in the athermal limit and with interfacial cohesion. In section 6.3, we study the
athermal limit and focus on the following issues: (1) The influence of static surface
corrugation of rough hard sphere (HS) fillers on their statistical spatial organization in a
dense polymer melt matrix. (2) The consequences of a fluctuating surface at a given static
surface corrugation. (3) The variation of static surface corrugation. (4) Comparison
between the incoherent fluctuation model and the coherent fluctuation model. Section 6.4
133
studies the role of interfacial cohesion and answers the following questions: (1) How
does surface fluctuation modify “bridging” phenomena given the same static corrugation?
(2) How does the filler PMF vary with particle size? (3) How can we make connections
with prior studies of smooth HS fillers? The chapter concludes in section 6.5 with a
summary and discussion. Section 6.6 and 6.7 present the references and figures,
respectively.
6.2 Effective Interactions
6.2.1 Nanoparticle-nanoparticle, ( )eff
nnU r
Under purely hard core conditions, we first examine the effective potential for the
frozen corrugated surface case, i.e. all beads are fixed at the particle surface. The
resulting interaction is a soft repulsion. When there is surface fluctuation, ( )eff
nnU r is
softer. Let us first examine its functional forms (See Fig.6.2 and 6.3), which is
qualitatively the same for all athermal cases. The softer the particles are (the larger the
surface fluctuation magnitude is), the softer the repulsive potential is, regardless of the
particle-monomer size ratio (only d σ= case is shown here.). The effective interaction
(Fig.6.2) starts from zero when the two outer shells (one can draw a circle that can
envelope the bumpiness formed by the surface beads) first touch each other. Then as the
two particles approach closer, the beads on one particle fit into the crevices of the other
particle’s surface, and a soft repulsion grows in. When particles can effectively feel the
full excluded volume geometric constraint, no smaller CM separation distance exists that
avoids overlaps, and ( )eff
nnU r diverges.
6.2.2 Nanoparticle-monomer, ( )eff
mnU r
134
We first show the athermal condition case. The effective interaction between one
particle and one monomer ( )eff
mnU r (see Fig.6.3) is qualitatively similar to ( )eff
nnU r . Since it
depends on monomer size, the repulsion result shifts to larger separations as monomer
size increases. The potential shape approaches the bare hard-core interaction as the
monomer size increases because the surface corrugation becomes less important if
monomer averages out the surface roughness (see Fig.6.4).
Then we show the case with interfacial cohesion. Let us look at the function form
of the effective interaction (see Fig.6.5). We denote the magnitude of the minimum of the
interaction as mn
ε , and the separation distance as mn
σ , “first contact distance”, which is
the smallest distance that a monomer can touche one surface bead along the radial
direction under frozen surface condition. We find the following features. For frozen
surfaces, mn mb
Sε ε= , where S is a constant depending on size asymmetry /D d . mn
ε
scales linearly with mb
ε , and the prefactor is a function of particle size asymmetry, i.e.
( / 5) 3.09S D d = = , ( / 7.8) 3.14S D d = = and ( / 10) 9.44S D d = = . The increase of this
prefactor is due to the change of particle surface curvature which induces more contacts
between surface beads and monomer. The length mn
σ remains the same for different
attraction strengths, ( ) / 2mn
D dσ σ= + + . Using mn
σ as a characteristic separation
distance, the right branch of the potentials collapses for different attraction strengths. This
means that at large separation distance, the particle level attraction range remains the
same for different attraction strength given the same local (bead level) attraction range,
surface corrugation does not matter. The left branch of the potentials splays apart because
of particle surface corrugation. For fluctuating surfaces, mn
ε does not scale linearly with
135
mbε , and there is only a minor increase as particle size increases (i.e. /D d increases
from 5 to 10). The first contact distance,mn
σ , ranges from above ( ) / 2D dσ+ + to below
this value as mb
ε increases. The right branch of the potentials collapses for different
attraction strengths while the left branch splays apart.
6.3 Athermal Systems: Pair Correlations & Thermodynamics
6.3.1 Frozen surface: Influence of static surface corrugation
In this case, the positions of the beads on the surface of nanoparticles are fixed,
thereby allowing study of the pure effect of having surface corrugation on hard fillers.
Fig.6.6 presents the particle-monomer pair correlation function. Since the
particles now have a rough surface, it is impossible for monomers to be in direct contact
with the core, thus it goes to zero strictly before close contact. (For MC, due to the
particle geometry, there will be a closest approaching distance pn
σ that is larger than
close contact distance ( ) / 2D d+ . Technically, this is determined when ( )pn
g r goes to
zero.) The position of the primary ordering peak is shifted to larger separation
accordingly.
The primary ordering peak is generally more pronounced for smooth HS fillers
than rough HS fillers. This is understandable because in the rough case, the monomer can
no longer touch the core directly. Instead, there is extra disorder when monomers explore
crevices, driven by the desire to maximize entropy. Thus, the primary ordering peak gets
broader, and the magnitude gets smaller. As one increases monomer size with particle
size fixed, the magnitude of primary peak for smooth HS decreases monotonically.
However, for rough HS cases, it increases monotonically. This is understandable since
136
when monomer size increases, the surface corrugation appears “smoother”, i.e. the
surface corrugation is averaged out. Thus, the larger the monomer is, the smaller the
difference in the magnitude of primary peak for the smooth and rough cases.
Fig. 6.7 presents the PMF between nanoparticles, which in general has two
origins. One comes from the direct interaction, the other comes from the polymer-
mediated interaction. Thus
( ) ( ) ( )eff polymer
nn nn nnW r U r W r= + (6.1)
The potential of mean force diverges at a larger interparticle separation for rough HS
cases because the surface corrugation introduces geometric constraint. The entropic
polymer mediated depletion attraction is weakened generally with surface corrugation.
This is understandable as the entropic penalty for polymers to pack around the surface of
nanoparticles is reduced as the monomers can now fit into the crevices of the surface
beads. As monomer size increases with particle size fixed, for smooth HS systems the
depletion attraction becomes smaller. For rough HS cases, the weakest depletion
attraction occurs for the largest monomer size, but the magnitude of depletion attraction
is similar for the other two cases while larger monomer case can feel the geometric
constraint at a larger particle separation.
The particle second virial coefficient, 2B ,characterizes the total thermodynamic
pair interaction between two nanoparticles and consists of both a “direct” and an
“indirect” (polymer-mediated) contributions.
( )
2,
1( 1)
2nn
W r
nnB dr e
β−= − −∫
(6.2)
137
To obtain the spinodal demixing condition for a dilute effective one-component (E1C)
system, one needs 2 2, 2,/nn reference
B B B= , where 3
2, 2, 2 / 3reference smHS
B B Dπ= = for smooth
HS fillers. However, this reference state is varying for rough hard spheres as the static
surface corrugation weakly changes, 2, 2,RoughHS smHSB aB= . The prefactor “a” is calculated
based on the rough HS pair potential between two nanoparticles.
( )( 1) / 2
eff
nnU r
a d r eβ−= − −∫
. As a result, the normalized 2B has taken into account of the
direct geometry difference due to surface corrugation.
We then analyze the 2B for frozen surface fillers. It is always negative for all
cases implying significant depletion attraction. The absolute value of 2B is always larger
for smooth HS than rough HS fillers, meaning that the surface corrugation reduces
depletion attraction. For smooth HS cases, as one increases monomer size with particle
size fixed, the absolute value of 2B monotonically decays while for rough HS cases the
change is non-monotonic. This is understandable because the smaller the monomers are,
the bigger influence the surface corrugation has on 2B . The crevices that monomers can
fit into reduce depletion attraction.
The spinodal solubility limit volume fraction, which is often used as a rough
indicator for particle dispersion, is given by [2]:
1
2(8 )s t
Bφ η −= − (6.3)
Surface roughness enhances dispersion by a factor ranges from 1 ( 2d σ= ) to 1000
( 0.5d σ= ) compared with its smooth hard sphere filler counterpart depending on size
asymmetry D/d (see below).
138
The experiment [1] which serves as one of our motivation found that it is possible
to disperse crosslinked polystyrene nanoparticles in linear polystyrene chain matrix. This
fact is determined from Small Angle Neutron Scattering (SANS) experiments via
presence or absence of fractal-like scattering at small wave vector and at nanoparticle
concentrations of 2 wt %. The fractal-like behavior is indicative of contact clustering
aggregates. Our theoretical calculation provides a possible explanation via the existence
of crevices formed by surface corrugation which can “mute” depletion attraction and
enhance dispersion.
6.3.2 Fluctuating surface: Influence of surface corrugation and fluctuation
Fig.6.8 shows the particle-monomer pair correlation function. Since the particles
now have a rough surface, it is impossible for monomers to be in direct contact with the
core, thus it goes to zero equal or larger than close contact distance ((D+d)/2=0.6D for
0.2d Dσ= = ). The primary ordering peak is most pronounced for the frozen surface
case. The larger fluctuation magnitude is, the less ordering is found as the primary
ordering peak is less pronounced. This is understandable because the monomer can
explore more space around the nanoparticle as the surface is soft and the corrugation is
not frozen. Also, for fluctuating surface cases, it goes to zero at smaller separation
distances approaching close contact distance.
139
Fig.6.9 presents the PMF between nanoparticles. When the surface “looks rough”
on the monomer size scale, i.e. d σ< (see Fig.6.9(A)), the depletion attraction gets
stronger both in terms of the attraction strength (the minimum) and spatial range,
compared to the frozen limit, with growing surface fluctuation. This can be understood in
the sense that surface fluctuation tends to average out surface corrugation, which is the
key factor to reduce the entropic penalty for polymers to pack around the nanoparticles.
However, when surface “looks smooth” on the monomer size scale, i.e. d σ≥ (see
Fig.6.9(B)), this effect becomes significantly less pronounced. For 2d σ= the minimum
of the potential goes up with increasing surface fluctuation magnitude.
For N=72 particles (see Fig.6.10), the 2B are negative for all surface fluctuation
amplitudes. In addition, increasing surface fluctuation magnitude leads to more
negative 2B , which is an indication of strong depletion aggregation. This conclusion is
most valid when the monomer size is equal or smaller than the corrugation size. When
the monomer size is larger than the corrugation scale, the effect can be reduced and the
trend can even be reversed. The spinodal solubility limit filler volume fraction, s
φ ,
follows directly from 2B (see Fig. 6.11(A) & (B)). Systems become less miscible as
surface fluctuation increases (see Fig.6.11(C)). Similarly, solubility is not sensitive to
surface fluctuation if the monomer size is larger than corrugation size. The system can
even be slightly more miscible as fluctuations get large. Compared with smooth HS
fillers, there will be a critical fluctuation magnitude, below and above which miscibility
can change several orders of magnitudes. In summary, the existence of surface
corrugation enhances dispersion, while for N=72 case, surface fluctuation suppresses
140
dispersion when the surface “looks rough” to the polymer monomers. However, is this
conclusion generally true?
6.3.3 Variation of static surface corrugation: Changing /D σ
The relevant surface corrugation that serves as an extra source of disorder by
creating crevices and weakens depletion attraction can be varied in several ways: (1) Fix
/d σ and change /D σ , i.e. change the particle size given the fixed local surface
corrugation size, or (2) fix /D d and change /D σ , i.e. change the static surface
corrugation size given the same particle size. The following studies of the particle second
virial coefficient and miscibility are carried out accordingly.
The matched bead-monomer size dσ = condition is perhaps most relevant to the
matched chemistry crosslinked polystyrene nanogels in linear polystyrene systems [].
However, why one might consider the monomer size to be the relevant corrugation size is
not straightforward to answer based on our coarse-grained models. The bead on the filler
surface is meant to mimic the most relevant length scale to discuss the effect of surface
corrugation on the depletion effect. As a result, although groups on the surface larger than
the monomer size scale do exist and contribute to surface corrugation, it is the monomer
length scale that matters for depletion effect and dispersion properties.
We now study multiple particle sizes ranging from / 5D σ = , 7.8 to 10 (the
corresponding numbers of beads are 72, 162, 282) for the same monomer size (bead size).
The trend is straightforward. (1) At zero fluctuation (see Fig.6.12), the larger the particle
is, the stronger the depletion attraction is. (2) For all particle sizes, as the fluctuation
magnitude increases, the depletion attraction also gets stronger. For smaller particles, the
attraction range increases while the attraction strength (the minimum of the PMF)
141
remains almost the same magnitude. In contrast, for large particles, both the range and
strength increase. Surface fluctuation suppresses dispersion despite the size differences
(see Fig.6.13). (3) However, for all cases studied, the rough fillers are more miscible than
their smooth HS counterparts. The dependence of the filler spinodal solubility limit
volume fraction on fluctuation magnitude follows roughly an exponential law. Our
physical understanding is that the monomers can only sense the local surface roughness.
Although the overall particle size is changed, which leads to the change of surface
curvature, the overall trend in terms of the dependence on fluctuation magnitude remains
the same, but the absolute value of solubility will depend on the particle-monomer size
asymmetry (typically grows strongly as an exponential), similar to the smooth HS cases
[2].
Based on previous studies of smooth HS filler polymer nanocomposites, we know
that the potential of mean force scales linearly with particle-monomer size asymmetry [2].
Thus,
( / )
( ) ( / )
2
1 1 1nn
A D d
s W r K D de
B dre dreβ β
φ −
−−
∫ ∫∼ ∼ ∼ ∼ (6.4)
and hence, we can roughly write the following relation:
,ln [ / ]s smHS
A D d Bφ = − + (6.5)
where A, B are two constants. By fitting the theoretical data (see Fig.6.14 squares) for
D/d=5, 7.8, 10 for smooth HS fillers, one has A=0.922, B=0.266. For soft and
interfacially rough nanoparticle fillers, the dependence on size asymmetry D/d is also
roughly exponential decay (see Fig.6.14).
,ln ( )[ / ] ( )s roughHS
A u D d B uφ = − + (6.6)
142
where A(u) and B(u) are now functions of the fluctuation magnitude, where A(u)<A and
A(u) increases with increasing fluctuation magnitude while B(u) decreases.
Given the above analytic function forms, we can quantify the enhancement of
solubility relative to the smooth HS filler cases via the ratio , ,/s s Rough s Smooth
φ φ φ= . The
relative enhancement for different nanoparticle sizes with matched bead-monomer size
can thus be compared:
,1 ,2 1 2ln ln ( ( ))[( ) / ]s s
A A u D D dφ φ− = − − (6.7)
Since A>A(u), then if 1 2D D> , one has ,1 ,2s sφ φ> , which is exactly the trend we observe
from numerical calculations (see Fig.6.15). Larger particles result in stronger
enhancements of dispersion compared to their smooth analogs. This difference decreases
as the fluctuation magnitude increases because A(u) is approaching A. Our physical
understanding is that there are two competing forces that affect dispersion. One is the
particle-monomer size asymmetry; the other is the bead-monomer size match. The larger
the particle is, the bigger the particle-monomer size asymmetry is, which reduces
dispersion. On the other hand, given the matched size condition, the larger the particle is,
the more crevices there are on the surface of the particle, which enhances dispersion. As a
result, when one normalizes the spinodal volume fraction by its value for smooth HS
filler cases, we are subtracting the effect of size asymmetry and leaving only the effect of
crevices.
We now consider the more general condition with mismatched bead-monomer
size. The starting point is to compare with smooth HS fillers of the same particle-
monomer size asymmetry.
143
First consider frozen surfaces. We have multiple corrugation sizes ranging from
/ 1dσ = , 0.7 to 0.5 (the corresponding numbers of beads are 72, 162, 282). The trend is
straightforward: At zero fluctuation (see Fig.6.16), the trend is as expected. As surface
gets smoother (corrugation size shrinks), it approaches smooth hard sphere limit. The
physical understanding is that surface looks smoother for the monomer as corrugation
size shrinks. (Literally, it goes to smooth HS as corrugation size vanishes in a extremely
slow manner.)
When surface fluctuation is present, the PMF displays complicated trends. At the
matched size condition, depletion attraction is enhanced as the fluctuation magnitude
increases, while the trend is reversed when beads become significantly smaller than
monomer size (see Fig.6.17). We can then normalize the spinodal volume fraction by its
value for smooth HS filler cases (see Fig.6.18)) and explore the whole regime of
mismatched size. As one varies the bead size from half the monomer diameter to twice
the monomer diameter, the dependence on fluctuation magnitude changes from slightly
increasing ( 0.5dσ = ) to dramatically decreasing ( 2dσ = ). The physical understanding
is that fluctuation makes the surface “look smoother”. This effect is most pronounced
when the surface is “rough” in the view of the monomer, which is when the bead size is
significantly larger than the monomer size and the monomer can fit into the crevices and
gain entropy.
6.3.4 Comparison: Incoherent fluctuation vs. coherent fluctuation
We have made an assumption about the way the surface beads fluctuate. It is the
incoherent radial Gaussian fluctuation. Coherent fluctuation means that all beads on one
particle fluctuation in the same phase, thus behaves effectively like “breathing”,
144
essentially an opposite limiting picture. For real crosslinked nanogels, the surface is a
crosslinked network with strands of monomers between nodal points. For small
fluctuations, the incoherent assumption should be most relevant. However, if fluctuation
gets significantly large, the connectivity between different monomers will be interfering
and tangential fluctuation will matter. This will also induce correlation between different
monomers. We are not trying to realistically address this effect in the present model. But
we can consider the extreme case, which is the coherent fluctuation model.
Fig.6.19 compares the filler PMF based on incoherent and coherent fluctuation
models. At small surface fluctuation, the coherent and incoherent calculations are close.
At large fluctuations, coherent effective interactions tend to be less repulsive. The
physical understanding is that how repulsive the inter-particle potential is largely depends
on the probability of overlapped configurations. Thus, at large fluctuation magnitude the
“dynamical disorder” generated by incoherent fluctuation would be most pronounced to
induce more overlapped configurations than the case at small fluctuations.
We now further investigate this effect in terms of particle-particle potential of
mean force (see Fig.6.20) and miscibility (see Fig.6.21(A) and (B)). Different from the
relatively weak dependence on fluctuation magnitude for incoherent fluctuation cases, the
coherent fluctuation models show much stronger dependence on fluctuation amplitude.
For the PMF, both the depletion attraction strength and range increase dramatically as the
fluctuation amplitude gets larger. This is a direct result of the interparticle potential,
where coherent fluctuation generates much less repulsion for large fluctuation. In terms
of miscibility, the coherent fluctuation cases can have solubility limits far below the
values for their smooth HS counterparts. We also study the dependence of miscibility on
145
particle size. Despite the differences in terms of fluctuation magnitude, the incoherent
fluctuation systems show a drop of miscibility of around one order of magnitude, while
the miscibility for coherent cases drops even more dramatically as fluctuation amplitude
gets large.
A tentative simple interpretation concerning the above results is that the solubility
of coherent bead fluctuation or breathing sphere might be understood by post-facto
averaging the smooth HS filler solubility (or second virial coefficient) over particle size
(D) fluctuation:
20
2
( )
(2 )( )
D D
uP D e
−−
∼ (6.8)
Thus we construct:
0
( / )
/ ,0
( ) ( )A D d
s D d uD e P D dDφ
∞−< > ∫∼ (6.9)
and the following equation can be derived and using method of steepest descents (valid if
/ 1D d >> ) to analyze its dependence on particle-monomer size asymmetry D/d:
220 0
220
2 20
220
( 1)( )( )
(2 /( / ))( / ) (2 )
0 0
( ) ( ) ( 1)
(2 /( / ))(2 )
0 0
00 0
0 00
( / , , )( / , )
( / , ) ( / , , )
DD D DA D
u D dA D d u d
s p D D D D
u D du
e e dD e e dD
e dD e dD
D d D u dDD d u
D d u D d D u dD
λφ λ
ωλ λ
ψ
−−−− − ⋅∞ ∞−
− −−−
∞ ∞
∞
∞
∫ ∫< > = =
∫ ∫
Ω ∫= =
Ψ ∫
(6.10)
where 0/D D D= and /u u d= , Then
* 2 2 2 2
0 0 0 0 0( / , ) ( / , 2 ( / ) 1, ) exp[ ( / ) ( / ) ]MAX
D d u D d D A D d u u A D d A D d uωΩ = = − + = − + (6.11)
*
0 0( / , ) ( / , 1, ) 1MAX
D d u D d D uψΨ = = = (6.12)
Thus,
146
2 20
2 20 0
( )0
( / , )ln[ ] ln[ ] ln[ ] ( ln )
( / , )
DA A u
d
s p D
D d u De A A u
D d u dφ λ λ λ
− +Ω< > = = = − + +
Ψ (6.13)
This simple analytic analysis qualitatively agrees with our numerical calculation (see
Fig.6.21(C)) which shows that on a log-linear plot, the slope for the averaged results are
the same as the smooth HS filler, but the intercept increases as fluctuation magnitude
squared. This behavior is the opposite of the results for the coherent fluctuation
calculations, which show that the miscibility decreases as fluctuation amplitude increases.
So far, we have compared the two models for the matched size case, we now
move to mismatched size cases. For the D/d=5 filler, we can also vary the bead size while
keep D and d constant (see Fig.6.22(A)). Despite the complicated trend observed for
incoherent fluctuation cases, the trend for coherent fluctuation is simple. The spinodal
solubility volume fraction drops as the fluctuation magnitude increases. This is least
pronounced when monomer size is larger than bead size. We can then normalize the
spinodal volume fraction by its value for smooth HS filler cases (see Fig. 6.22(B))) and
explore the whole regime of mismatched size. The trend is similar to the one observed for
the incoherent fluctuation cases but is more pronounced.
6.4 Role of Interfacial Cohesion
6.4.1 Particle-monomer pair correlation function
Fig.6.23 shows some examples of the particle-monomer pair correlation function
for N=72 particle ( / 5D σ = ) with matched size ( d σ= ) and various attraction strengths.
We study both the frozen and fluctuating ( 0.25u σ= ) cases. One can define a closest
approach distance as the separation at which the pair correlation function vanishes.
Physically, this is the smallest center of mass separation distance between a monomer and
147
a nanoparticle. i.e. ( ) / 2pn
D dσ = + for smooth HS filler system and rough fluctuating
surface nanoparticle system because monomers can touch particle core while
( ) / 2pn
D dσ > + for rough, frozen surface nanoparticle system, for which there is surface
geometry constraints due to corrugation. e.g. 0.636pn
Dσ = for D/d=5 system where
( ) / 2 0.6D d D+ = . As surface attraction strength grows, the primary ordering peak shifts
to smaller separation distance and has a larger magnitude, indicating a selection of length
scales corresponding to maximum adsorption of monomers on nanoparticle surface. The
stronger the attraction strength is, the more adsorption onto the particle surface occurs.
Surface fluctuation competes against this length scale selection by making the primary
peak broader and less pronounced. This effect is stronger as interfacial attraction
increases.
6.4.2 Particle-particle potential of mean force
We investigate the impact of surface corrugation and fluctuation on the local
“bridging” of fillers by polymer chains by answering the following three questions:
(1) How does surface fluctuation modify “bridging” phenomena given the same
static corrugation? Let us look at Fig.6.24(A). Here we present the PMF for fixed surface
static corrugation / 5D σ = (N=72). We find that no matter what regime one studies, i.e.
contact clustering, steric stabilization or bridging, surface fluctuation induces more
attraction in the same manner as in the athermal limit. The underlying physics is common,
fluctuation smears out surface corrugation and reduces interfacial cohesion, leading to
stronger depletion attraction.
(2) How does the PMF vary with particle size? Based on the answer to question
(1), we can answer this by studying frozen surfaces. In Fig.6.24(B) results for two
148
particle sizes are shown for similar attraction strength both in the local bead-monomer
level and particle-monomer level. Similar as under athermal conditions, the larger the
particle is, the stronger attraction the PMF shows. This is due to the size asymmetry
induced depletion attraction.
(3) How can we make connections with the prior studies of smooth HS fillers? In
smooth HS filler PNCs, there are only three length scales D, d and α , as well as one
energy scale ε . In frozen, rough filler PNCs, there are four length scales D, d, σ and
mbα , as well as one local energy scale
mbε . There is no unique way to map the entire
interfacial attraction profile to smooth HS cases where only α andε matter. This is due to
the active involvement of new length scales in the problem associated with surface
corrugation and fluctuation. Using the “first contact distance” mn
σ as a borderline, the
effective interaction between one particle and one monomer ( )eff
mnU r can be divided into
left (mn
r σ< ) and right (mn
r σ> ) branches. When we change the local bead-monomer
attraction strength mb
ε , mn
ε and the left branch will be modified. However, the right
branch (or literally, the attraction branch) almost remains the same, meaning the
“attraction range” on the particle-monomer level remains the same despite the variation
of the local bead-monomer attraction range, i.e. mn
constα α≈ = . For example, N=72 and
D/d=5, and 0.5mb
dα = , we find 0.546mn
dα α≈ = . However, this constant α changes
with particle size: for N=162 and D/d=7.8, 0.5mb
dα = and 0.558mn
dα α≈ = ; for N=282
and D/d=10, 0.5mb
dα = and 1.245mn
dα α≈ = . We then can make two possible
connections or “mappings” to smooth HS filler calculations (see Fig.6.24(C)):
149
(1) Ignore the effect of the left branch and replace this with a hard core interaction
and shift the entire potential profile so that the hard core interaction enters at
( ) / 2pn
D dσ = + , i.e. we map the frozen surface ( , , )eff
mn mb mbU r ε α to ( , , )
pn mn mnU r ε α and
compare the PMF of the latter case with the frozen surface calculations.
(2) Use thermal energy to determine a new length scale eff
σ and carefully choose
an attraction range eff
α such that the total cohesion energy is the same as
( , , )eff
mn mb mbU r ε α (
eff mnα α> since it also accounts for the contribution from the left branch)
and shift the entire potential profile so that the hard core interaction enters at
( ) / 2pn
D dσ = + , i.e. we map the frozen surface ( , , )eff
mn mb mbU r ε α to ( , , )
pn mn effU r ε α .
Based on the two mapping methods described above, one can compute the
potential of mean force for the rough spheres and their mapping counterparts based on the
smooth HS filler model and compare them. No matter what mapping method we apply,
we can always recover almost all features given there is a shifted length scale due to the
ambiguity of defining closest contact distances. Despite the somewhat complicated nature
of such mapping procedure, the more detailed mapping (i.e. using mapping 2 instead of 1)
we apply to the potential, the better agreement in terms of the PMF between smooth and
rough sphere mapping can be achieved. (see Fig.6.24(D))
6.4.3 Second virial coefficient
Fig.6.25(A) shows that in the intermediate attraction strength, steric stabilization
regime, the dimensionless second virial coeeficients are not sensitive to the surface
fluctuation amplitude. However, for both the contact clustering and strong bridging
regimes, 2B strongly depends on fluctuation magnitude and decreases (more attractive)
150
with increasing fluctuation amplitude. Fig.6.25(B) further confirms the above conclusion
by showing a strong non-monotonic dependence on attraction strength that ranges from
contact clustering ( 0mb
βε = ) to steric stabilization ( [0.25,0.75]mb
βε ∈ ) to strong
bridging ( 1mb
βε = ).
6.5 Summary and Discussion
In this chapter, we have presented representative results based on a new
minimalist multi-scale model for soft nanoparticle fillers. This is the first theoretical
study of the role of nanoparticle morphology associated with surface corrugation and
fluctuation at one and two particle limit in polymer melts with/without interfacial
cohesion.
Our results provide a physical basis for the unexpected ability to disperse
chemically matched crosslinked polystyrene in linear polystyrene melt. The surface
corrugation in the frozen surface regime results in a favorable entropic driving force for
mixing which competes with unfavorable depletion, resulting in a major enhancement of
nanoparticle dispersion, including a much larger spinodal solubility limit. Smooth hard
sphere behavior is recovered when the number of beads are significantly large (N=282)
and the relative corrugation size significantly reduced (from 20% to 10% of particle size).
When surface fluctuation exists, the dependence of solubility limit on fluctuation
magnitude is somewhat subtle, due to the relevance of multiple length scales. We find
fluctuation suppresses dispersion for all particles sizes (or surface curvature) studied, and
this effect is most pronounced when the monomer size is smaller than corrugation size.
The extreme coherent fluctuation model shows dramatically enhanced dependence on
151
fluctuation magnitude at large fluctuation magnitude and less miscibility. When
interfacial cohesion exists, we develop a model for local bead-monomer level attraction,
which explicitly connects to particle-monomer level attraction through potential mapping
strategies. By varying the interfacial attraction strength, we can still observe all three
regimes reported in prior PRISM studies of smooth HS filler PNCs. The steric
stabilization regime is not sensitive to surface fluctuation magnitude, while the contact
clustering and strong bridging regimes are very sensitive to surface fluctuation magnitude.
Surface fluctuation smears out surface corrugation and reduces interfacial cohesion,
leading to stronger depletion attraction and more negative second virial coefficients.
Recent SANS experiments show that the dispersion of crosslinked polystyrene
nanogels (with diameters that range from ~5 nm to 7 nm or larger) in chemically matched
dense polymer matrix of linear chain-like polystyrene (monomer size~1 nm) is possible [].
We know this not true if the filler is smooth HS with no interfacial cohesion. So it is now
natural for us to believe the surface corrugation and fluctuation plays a major role to
achieve this “high” dispersion (nanoparticle concentration of 2 wt %). Our multi-scale
model explicitly studies the surface corrugation and fluctuation and our hybrid approach
proves the possibility to achieve dispersion and predicts the necessary conditions required
(large surface corrugation and small surface fluctuation). Our predictions of the second
virial coefficient can be tested via dilute filler SANS experiments, which are in progress
at Oak Ridge National Lab. Apart from experiments, full computer simulations (using our
multi-scale modeling) of such soft, interfacially rough nanoparticle PNCs can also be
designed to test the major results reported in this chapter, e.g. the validity of incoherent
152
fluctuation model versus coherent fluctuation model, the reduction of miscibility as
surface fluctuation increases, the dependence of miscibility on multiple length scales, etc.
This work presented here was performed in the dilute particle limit. A natural
extension of this hybrid small scale MC simulation plus PRISM theory approach is the
determination of the mixture structure and the spinodal phase diagram beyond the low
filler volume fraction regime. Prior PRISM studies have found qualitatively new insights
concerning structure and phase behavior, e.g. the identification of critical points and
construction of full spinodal curves, from the dilute filler to dilute polymer limits [3]. The
many body particle correlation effects are perhaps important for understanding
miscibility when the mixture has a positive filler second virial coefficient and repulsive
dilute filler limit potential of mean force. This is an indication that the system can
undergo a collective bridging driven phase separation [3]. Fourier space partial collective
density fluctuations are experimentally measurable using selective labeling scattering
methods, which can determine the system miscibility from dilute to ultrahigh filler
volume fractions. [4]
The present work serves as a starting point for the development of a predictive
theory of multiple phenomena in real soft filler polymer nanocomposites, such as
crosslinked nanogels, chemically manipulated colloidosomes, and also motivates further
generalization of NMCT and NLE theories [5-7] to treat the role of surface morphology
for slow dynamics of soft colloid suspensions like crosslinked microgels (as discussed in
next chapter) and micelles formed by block copolymers, which also has an impenetrable
core and an outer soft coronal layer.
153
6.6 References
[1] M. E.Mackay, A. Tuteja, P.M.Duxbury, C. J.Hawker, B Van Horn, Z. Guan, G Chen,
R. S. Krishnan, Science 311, 1740 (2006)
[2] J. B. Hooper and K. S. Schweizer, Macromolecules 39, 5133 (2006).
[3] L. M. Hall and K. S. Schweizer, J. Chem. Phys. 2008, 128, 234901.
[4] B. Anderson and C. F. Zukoski recent experiments
[5] T.R.Kirkpatrick and P.G.Wolynes, Phys. Rev.A, 35, 3072 (1987).
[6] K.S.Schweizer and E.J.Saltzman, J.Chem.Phys., 119, 1181(2003).
[7] K.S.Schweizer, J.Chem.Phys., 123, 244501(2005).
154
6.7 Figures
Fig. 6. 1
Conceptual scheme: A schematic representation of the hybrid simulation-theory approach
to study soft, interfacially rough and fluctuating filler polymer nanocomposites. We start
from the smooth hard sphere filler, then add surface beads to model static surface
corrugation. The static surface corrugation can be varied by changing the number of
beads on the surface. Surface fluctuation is then added by allowing beads to fluctuate
radially either in an incoherent or coherent manner to vary the “dynamic” surface
corrugation.
155
Fig. 6. 2
Effective pair potential between nanoparticle and monomer are shown for various
fluctuation magnitudes. Lines indicate polynomial fitting results.
156
Fig. 6. 3
Effective pair potential between two nanoparticles are shown for various fluctuation
magnitudes. Lines indicate polynomial fitting results.
157
Fig. 6. 4
Effective pair potential between two nanoparticles and between one particle and one
monomer is shown for various D/d at fixed / 5D σ = (N=72 particle). Open symbols
indicate the cross monomer-filler pair potentials. We vary D/d =10, 5, 2.5 by varying
/ dσ = 0.5, 1, 2.
158
Fig. 6. 5 (A)
The effective pair potential between nanoparticle (N=72) and monomer for frozen and
fluctuating rough surfaces is shown as a function of particle displacement for various
attraction strengths.
159
Fig. 6.5 (B)
The effective pair potential between nanoparticle (N=282, D/d=10) and monomer for
frozen and fluctuating rough surfaces is shown as a function of particle displacement for
various attraction strengths.
160
Fig. 6. 5 (C)
The effective pair potential between nanoparticle and monomer for frozen and fluctuating
rough surface non-dimensionalized by the corresponding local attraction strength are
shown as a function of particle displacement.
161
Fig. 6. 5 (D)
The effective pair potential between nanoparticle and monomer for frozen and fluctuating
rough surface non-dimensionalized by the corresponding local attraction strength are
shown as a function of particle displacement.
162
Fig. 6. 6
Particle-monomer pair correlation function for the frozen suraface is shown as a function
of particle-monomer separation at fixed / 5D σ = (N=72 particle) for various D/d. We
vary D/d =10, 5, 2.5 by varying / dσ = 0.5, 1, 2.
163
Fig. 6. 7
Nanoparticle potential of mean force for the athermal case is shown for various D/d.
Smmoth hard sphere fillers calculations are also shown for reference.
164
Fig. 6. 8
Nanoparticle-monomer pair correlation function is shown as a function of separation
distance for various fluctuation magnitudes for D/d=5.
165
Fig. 6. 9 (A)
Nanoparticle potential of mean force for D/d=10 is shown as a function of particle
separation distance for various fluctuation magnitudes. The monomer size is chosen to be
half the surface corrugations size. Smooth HS result for the same size asymmetry
(D/d=10) is also shown for comparison.
166
Fig. 6. 9 (B)
Nanoparticle potential of mean force for D/d=5 (solid) and D/d=2.5 (dash) is shown as a
function of particle separation distance for various fluctuation magnitudes. The monomer
size is chosen to be equal to or twice the surface corrugations size. Smooth HS results for
the same size asymmetry (D/d=5, 2.5) are also shown for comparison.
167
Fig. 6.10
Negative of the dimensionless second virial coefficient is shown as a function of bead
fluctuation amplitude for N=72 particle and various monomer sizes. Smooth HS
calculation for the same size asymmetry is also shown for reference.
168
Fig. 6. 11 (A)
Spinodal solubility limit volume fraction is shown as a function of beads fluctuation
magnitude for N=72 particle and various monomer sizes. Smooth HS calculation for the
same size asymmetry is also shown for reference.
169
Fig. 6. 11 (B)
Spinodal solubility limit volume fraction is shown as a function of bead fluctuation
amplitude for N=72 particles and various monomer sizes. Smooth HS calculation for the
same size asymmetry is also shown for reference.
170
Fig. 6. 11 (C)
Spinodal solubility limit volume fraction normalized by its smooth hard sphere value of
the same size asymmetry is shown as a function of bead fluctuation amplitude for N=72
particles with varying monomer sizes.
171
Fig. 6. 12
Comparison of PMF for fluctuating spheres with /D σ =5, 10 for various fluctuation
magnitudes.
172
Fig. 6. 13 (A)
Spinodal solubility limit volume fraction as a function of bead fluctuation magnitude for
N=72, 162 and 282 particle. Smooth HS calculation for the same size asymmetry is also
shown for reference.
173
Fig. 6. 13 (B)
Spinodal solubility limit volume fraction as a function of bead fluctuation magnitude for
N=72, 162 and 282 particle. Smooth HS calculation for the same size asymmetry is also
shown for reference.
174
Fig. 6. 14
Spinodal solubility limit volume fraction as a function of particle-monomer size
asymmetry for various fluctuation magnitudes. Smooth HS calculation is also shown for
reference.
175
Fig. 6. 15
Spinodal solubility limit volume fraction normalized by its smooth hard sphere value of
the same size asymmetry is shown as a function of bead fluctuation magnitude for
d σ= with varying nanoparticle sizes.
176
Fig. 6.16
Comparison of nanoparticle PMF for rough hard spheres with /D σ =5, 7.8, 10 are shown.
Smooth hard sphere calculation is also presented for reference.
177
Fig. 6. 17
Nanoparticle PMF for D/d=5 with various surface corrugation sizes and fluctuation
magnitudes as a function of particle displacement.
178
Fig. 6. 18
Spinodal solubility limit volume fraction normalized by its smooth hard sphere value of
the same size asymmetry as a function of bead fluctuation magnitude for various bead-
monomer size ratios.
179
Fig. 6. 19
The effective pair potential between two nanoparticles for the incoherent and coherent
surface fluctuation models as a function of particle displacement.
180
Fig. 6. 20
Nanoparticle potential of mean force for D/d=5 as a function of particle separation for
various fluctuation magnitudes for both the incoherent and coherent fluctuation models.
181
Fig. 6. 21 (A)
Spinodal solubility limit volume fraction normalized by its smooth hard sphere value of
the same size asymmetry as a function of bead fluctuation magnitude for N=72, 162 and
282 particles. Results for both the incoherent and coherent fluctuation models are shown.
182
Fig. 6. 21 (B)
Spinodal solubility limit volume fraction as a function of size asymmetry for smooth HS
and rough HS and various fluctuation magnitudes. Dashed lines indicate the coherent
counterparts for the corresponding fluctuation magnitudes.
183
Fig. 6. 21 (C)
Spinodal solubility limit volume fraction as a function of size asymmetry for smooth HS
and rough HS and various fluctuation magnitudes. Solid lines indicate results for the
breathing smooth sphere models and dashed lines indicate the rough counterparts for the
corresponding fluctuation magnitudes.
184
Fig. 6. 22 (A)
Spinodal solubility limit volume fraction as a function of bead fluctuation magnitude for
D/d=5. Open symbols represent incoherent fluctuation cases, while solid symbols
represent coherent fluctuation cases. Smooth HS calculation for the same size asymmetry
D/d=5 is also shown for reference.
185
Fig. 6. 22 (B)
Spinodal solubility limit volume fraction normalized by its smooth hard sphere value of
the same size asymmetry as a function of beads fluctuation magnitude for various bead-
monomer size ratios and the coherent surface fluctuation model.
186
Fig. 6. 23
Nanoparticle-monomer pair correlation function for D/d=5 as a function of reduced
separation for various surface attraction strength (black, red, green, blue: 0, 0.25, 0.75, 1)
for frozen surface fillers. Smooth HS filler result (dash) is also shown for comparison.
Inset: The same calculations for fluctuating surface fillers ( 0.25u σ= )
187
Fig. 6. 24(A)
Nanoparticle potential of mean force for D/d=5 is shown as a function of particle
separation for various attraction strength and fluctuation magnitudes.
188
Fig. 6. 24 (B)
Nanoparticle potential of mean force for D/d=5 and 7.8 frozen surface model as a
function of particle separation for various attraction strengths.
189
Fig. 6. 24 (C)
The effective interfacial attraction is shown as a function of particle displacement. The
two smooth HS mappings are also shown for comparison. Mapping one (red): Hard core
interaction at ( ) / 2pn
D dσ = + plus an exponential attraction with an attraction range
mnα extracted from the right branch of ( )eff
mnU r . Mapping two: Hard core interaction at
( ) / 2pn
D dσ = + plus an exponential attraction with an attraction range eff
α extracted by
maintain the total cohesion energy, i.e. taking into account the effect of the left branch of
( )eff
mnU r .
190
Fig. 6. 24(D)
Nanoparticle potential of mean force for D/d=5 frozen surface (rough HS) as a function
of particle separation for various attraction strengths. The two smooth HS mappings are
also shown for comparison.
191
Fig. 6. 24(E)
Nanoparticle potential of mean force for D/d=5 frozen surface (rough HS) as a function
of particle separation for various attraction strengths. The smooth HS mapping is also
shown for comparison.
192
Fig. 6. 25(A)
Dimensionless second virial coefficient as a function of bead fluctuation magnitude for
N=72 particle and various attraction strengths.
193
Fig. 6. 25(B)
Dimensionless second virial coefficient as a function of attraction strength for N=72
particle and various bead fluctuation magnitudes.
194
Fig. 6. 25(C)
Dimensionless second virial coefficient as a function of attraction strength for N=72 and
N=162 particles with frozen surfaces.
195
CHAPTER 7
MULTI-SCALE MODELING OF
SOFT COLLOID MELTS
7.1 Introduction
Novel polystyrene nanoparticles were synthesized by the controlled
intramolecular crosslinking of linear polymer chains to produce well-defined single-
molecule nanoparticles of varying molecular mass and particle sizes. [1] These
nanoparticles are ideal to investigate the relaxation dynamics and rheology of high
molecular mass polymer melts in the absence of chain entanglements. Kratky plots
obtained from Small Angle Neutron Scattering (SANS) data show a shift toward particle-
like nature with increasing molecular mass and crosslink density. The viscoelastic
behavior of the particles in a 1-component liquid state was found to be strongly
dependent on both the extent of intramolecular crosslinking and molecular mass
(correlates with particle size), with a colloidal gel-like behavior at low frequencies
evident for a large degree of crosslinking where the nanogels are compact. The low
frequency elastic shear modulus, G’, grows as nanoparticle size, D, increases, in contrast
to the well established behavior for hard colloids, where flocculated suspensions exhibit a
modulus that scale inversely to the particle radius raised to a power of typically 2 or so
[2,3]. For jammed particle systems at zero temperature, a scaling inversely proportional
to the radius raised to the power of the system dimensionality [4] is present, i.e.
3' /B
G k T D∝ . These and other results suggest the presence of a secondary non-chain-
196
like mode of polymer dynamics relaxation, which is influenced by the total number of
crosslinks in the nanoparticle.
Our perspective in applying the methods developed in Chapter 5 to dense
suspensions of crosslinked nanogels is that although we still rely on center of mass level
interparticle potentials, the multi-scale modeling enables us to explicitly study the role of
particle size and crosslink density with regards to surface fluctuation or softness. This
variable is buried in the models of many arm stars or Hertzian spheres because there is
only one length scale, overall size. A melt or suspension of nanogels is an ideal system to
utilize the two-length-scale bead-core model because the internal degrees of freedom
indeed is not as relevant as the surface corrugation in the context of interparticle
interaction.
This chapter is focused on the slow dynamics of dense soft colloidal suspensions
based on the multi-scale modeling approach constructed in previous two chapters.
Section 7.2 discusses the model, theories and definition of effective volume fraction. A
study of the effect of particle size and surface fluctuation on mean alpha relaxation time
and the kinetic arrest map is presented in section 7.3. An application of NLE theory of
mechanical response to study the effect of particle size and surface fluctuation on linear
shear modulus is established in section 7.4. The chapter concludes in section 7.5 with a
summary and discussion. Section 7.6 and 7.7 present the references and figures,
respectively.
7.2 Model, Theory and Effective Volume Fraction
197
7.2.1 Multi-scale modeling, effective interaction and connections to real
crosslinked nanogels
We apply the theoretical approach developed in Chapter 5 to model soft,
interfacially rough nanoparticles with corrugated and incoherent fluctuating surfaces. We
use the same techniques used in Chapter 6 to calculate the effective interactions at center
of mass level to mimic the interaction between two soft nanoparticles. Fig.7.1 presents
representative potentials for the frozen and most fluctuating surfaces studied for three
particle sizes at fixed static surface corrugation. All cases show soft repulsive interactions
and surface fluctuation makes it even softer. Large nanoparticles interact more like hard-
sphere than small nanoparticles, i.e. the repulsion increases faster as one decreases the
separation distance.
A natural question one may have is why we are doing this calculation given that
we already performed calculations based on center of mass level interparticle potentials
like Hertzian model? The answer is we indeed used Hertzian model at the center of mass
level to capture the soft jamming, slow dynamics, and elasticity. But that model is based
on viewing a microgel as an elastic smooth sphere. This renders it is impossible to a
priori study the particle size dependence given it is the only length scale in the model
which is subsumed in the volume fraction. This stands in contrast to real nanogels where
there are site (monomer) level length scales. However, the Hertzian model does account
for the fact that these nanoparticles are soft and deformable spheres and thus all “soft
jamming” related properties appear to be qualitatively correctly captured. In this chapter,
we take one step further towards a microscopic model, and view the nanogels as spheres
with rough and fluctuating surfaces. We are not treating all sites level lengths scales due
198
to its complexity, but rather focus on the nanoparticle size and surface corrugation length
scale which is directly related to interactions due to contacts, a key characteristic for
dense suspension of microgels/nanogels.
7.2.2 Naïve MCT, NLE and Effective Volume Fraction
As discussed in Chapters 2, 3 and 4, suspensions of repulsive hard and soft
spheres tend to kinetically arrest (vitrify) at very high volume fractions due to the local
cage effect. This phenomenon has been theoretically addressed based on ideal mode
coupling theory (MCT), and the beyond MCT barrier hopping nonlinear Langevin
equation (NLE) approach. Glass formation is a consequence of strong interparticle
ordering on the local cage scale. The signature of the latter is well developed solvation
shells in the radial distribution function, or a large enough cage scale peak in the static
structure factor.
Viewing the soft, interfacially rough spherical particles as effective colloids at the
CM level, one can ask if they also form glasses, or at least show a dynamic crossover (the
ideal MCT ‘‘transition’’) to transient localization and slow activated barrier hopping? We
have explored this question by applying the well-documented CM version of naive MCT
(NMCT) [5]. This methodology has been shown to accurately capture the soft jamming
transition, slow activated dynamic and elastic properties for dense fluids of Hertzian
spheres [6,7].
To connect with experiment and take into account the existence of geometric
constraints due to bumpy particle surfaces requires some idea how the volume of one
object can be defined. Moreover, the fluid volume fraction can take on values in excess
of the hard sphere random close packing value due to soft repulsive particle overlap. Two
199
‘‘local’’ length scales of a single particle follow directly from the hard core aspects of the
interactions: (1) surface bead diameter,σ and (2) core diameter, D.
For a fluid of such particles, there are multiple ways to define volume fraction, all
motivated, more or less, by an effective diameter idea in analogy with the hard sphere
fluid. In colloid science, the interparticle separation (eff
D ) at which the effective
interparticle potential equals zero or thermal energy is often used to define an effective
diameter, and based on the latter an effective volume fraction:
31
6eff c eff
Dφ ρ π= (7.1)
Despite the ambiguities of choosing the criteria to define the effective particle diameter
[8], we find our conclusions are not qualitatively sensitive to the precise choice.
7.3 Effect of Particle Size and Softness on Mean Alpha Relaxation Time and Kinetic
Arrest
Using Eq.(2.8) the mean barrier hopping time is computed as a function of surface
fluctuation magnitude and volume fraction. All times are expressed in terms of the
elementary short Brownian time scale,s
τ . To make quantitative comparisons with
experiment, we need to estimate the time unit, the elementary short Brownian time
scale,s
τ . This quantity is influenced by the complicated polymeric nature of nanogels. If
we believe this short Brownian time corresponds to a polymer chain equilibrium time
scale for the internal degrees of freedom to relax, then we can write:
2 /s s
D Dτ = (7.2)
200
where s
D corresponds to the short time diffusion constant and D is nanoparticle size.
Based on a Rouse model of unentangled polymer dynamics as a crude estimate of the
global shape (confirmation) equilibration relaxation time, then
2
0R WNτ τ= (7.3)
2 2
2
06 6s
R W
D DD
Nτ τ∼ ∼ (7.4)
where R
τ is the Rouse time, W
N is the number of monomers in a chain, and 0τ is the
elementary Brownian time for monomers. Given the temperature for the experiments is at
T~440K, one has 7
0 10τ −∼ sec as the segment relaxation time. For polystyrene with a
molecular weight of 100 kDa, given the styrene molar mass is 104.15 g/mol,
1000W
N = 1. As a result, the elementary short Brownian time is of order 1 sec.
Fig.7.2 presents the dimensionless mean hopping times as a function of volume
fraction for a range of particle surface fluctuation magnitudes (u). Results are shown over
a wide range of relaxation times (7 orders of magnitude, as relevant to typical
experiments and simulations.). In real colloidal suspensions (and often simulations too), a
relaxation time range of only typically 4-6 orders of magnitude can be probed, perhaps up
to 2 5/ ~ 10 10hop s
τ τ − (roughly 100 sec to 1 day) which represents a practical threshold
used to define kinetic vitrification consistent with the limitations of low frequency
rheometers. Fig.7.2 shows the relaxation times generally grows with eff
φ , and more
weakly as they soften, i.e. u increases.
1
5 27 233
3
( )( ) 10 1.66 10 6.022 1010
( ) ( ) 104.15 10
A
W
MolecularWeight PS NMolarMass PSN
MolarMass S MolarMass S
−
−
× × × × ×= = = ≈
×
201
As the particle size increases at fixed local corrugation, the relaxation time also
increases. In addition, the increment of relaxation time is getting larger as the fluctuation
magnitude increases. We use / 10X
hop sτ τ = (X=2, 3, 4, 5) as a practical threshold to
define kinetic vitrification volume fraction and construct the kinetic arrest map. As a
result, at the same relaxation time threshold, the bigger particle undergoes a kinetic
vitrification at a lower volume fraction. This trend of easier solidification for large
particles is qualitatively consistent with the polystyrene nanogel melt experiments [1].
Moreover, the more fluctuating the surface is, the higher volume fraction the system is
required to be kinetically arrested.
7.4 Effect of Particle Size and Softness on Linear Shear Modulus
The linear elastic shear modulus of the transiently localized system, G’, is
computed using a standard Green-Kubo-like statistical mechanical formula (Eq.(2.12) ).
Fig.7.3(A) presents calculations of G’, in units of 3
72Bk TD
− , as a function of effective
volume fraction for 3 particle sizes and over a wide range of particle surface fluctuation
magnitude, u. The growth of the modulus with effective volume fraction does not display
any simple functional form over a wide range.
To make qualitative contact with nanogel experiments, based on our
/ 5,7.8,10D σ = model systems (so D ≈ 5, 7.8, 10 nm for a polystyrene-like ~σ 1 nm),
we note that the theory predicts that larger particle has a higher elastic modulus, agreeing
with the nanogel melt experiment [1]. This modulus increase with soft particle size is
considered to be related to primarily one factor [1]. The larger molecular weight (larger
particle size) system is more ‘‘particle-like’’ in nature based on SANS experiments. We
202
believe that the more “particle like” can be interpreted as more “smooth hard sphere” like.
The interaction range decreases, with harshness of the repulsive force growth as two
nanogels collide, as particle size increases. This is physically understood in our model
because if the corrugation size is fixed, then increasing particle size means less
importance lies in the surface corrugation and thus more smooth-hard-sphere like
behavior.
We can make quantitative estimations of G’. Take 72 5D nm= and temperature
T=443K (reported particle size and temperature in [1]) and 310hop
τ = sec to define a
kinetic glass, we can replot Fig.7.3 (A) in absolute unit. Based on Fig.7.3 (B), by
enforcing the theory to recover the low frequency plateau value of elastic modulus for
tightly crosslinked nanogels: 410 Pa for 5nm particle, we find a volume fraction ranges
(depending on the fluctuation magnitude, larger fluctuation magnitude corresponds to
higher volume fraction) from 52% to 64%; and 43 10× Pa for 7nm particle, we find a
volume fraction ranges from 47% to 55%. These volume fractions are typical for melt
condition, so the interpretation seems plausible.
7.5 Summary and Discussion
In this chapter, we have applied the multi-scale model for soft nanoparticle
developed in Chapter 5 and 6 to study the role of nanoparticle surface corrugation and
fluctuation on the slow dynamics of dense suspensions. This methodology allows us to
study the dependence on particle size, which can not be explicitly investigated in the
simple models like Hertzian spheres [9] where only one length scale exists.
203
Our results provide a physical basis for the striking phenomena observed in the
experiment of crosslinked polystyrene nanogel melts (with diameters that range from ~3
nm to 7 nm or larger). (1) Larger nanogel melts form soft solids easier and have larger
low frequency shear modulus than smaller particles. This “counterintuitive” based on
classic hard colloids is the consequence of two competing factors: the scale of stress
storage 3/B
k T D decreases with particle size but the “hardness” of interparticle repulsion
increases.
We plan to combine the current advance with mechanical response theories to
describe the shear rheology of soft glasses from our multi-scale perspective. Although
polymer entanglement, osmotic deswelling etc. are not included in the model, it is still
the first of its kind study that encodes the basic aspects of surface corrugation and
fluctuation and extreme limiting cases of dynamic surface fluctuations. In particular, the
Maxwell viscosity, defined as the product of elastic shear modulus and mean barrier
hopping time (see Fig.7.4), shows an even more pronounced dependence on particle size.
This is a direct result of the combination of the enhanced dependence on particle size for
elastic modulus and alpha relaxation time.
7.6 References
[1] A. TUTEJA, M.E. MACKAY, C. J. HAWKER, B. VAN HORN, D. L. HO, Journal
of Polymer Science: Part B: Polymer Physics
[2] Russel, W. B.; Saville, D. A.; Schowalter, W. R. Colloidal Dispersions; Cambridge
University Press: Cambridge, 1989.
204
[3] Shah, S. A.; Chen, Y. L.; Schweizer, K. S.; Zukoski, C. F. J Chem Phys 2003, 119,
8747– 8760
[4] O’Hern, C. S.; Silbert, L. E.; Liu, A. J.; Nagel, S. R. Phys Rev E: Stat Phys Plasmas
Fluids Phys Relat Interdiscip Top 2003, 68, 011306
[5] T.R.Kirkpatrick and P.G.Wolynes, Phys. Rev.A, 35, 3072 (1987).
[6] J. Yang and K. S. Schweizer, J. Chem. Phys. 134, 204908 (2011).
[7] J. Yang and K. S. Schweizer, J. Chem. Phys. 134, 204909 (2011).
[8] There is no unique way to define the effective diameter. We choose the thermal
energy condition for all calculations but we also tried other definition, e.g. the separation
distance where the potential is equal zero. No qualitative difference exists.
[9] J.R.Seth, M.Cloitre and R.T.Bonnecaze,J.Rheology, 50, 353 (2006); 52, 1241 (2008).
205
7.7 Figures
Fig. 7. 1
Effective pair potentials between two nanoparticles are shown as a function of particle
separation distance for a frozen corrugated surface and fluctuating surface ( 0.25u σ= i.e.
2( ) 0.17r R σ< − > = see Chapter 5) for three particle core sizes of the same surface
corrugation size.
206
Fig. 7. 2
Dimensionless hopping time is plotted against effective volume fraction for various
particle sizes and fluctuation magnitudes. The four horizontal lines indicate four kinetic
glass transition criteria.
207
Fig. 7. 3 (A)
Dimensionless elastic modulus is plotted against effective volume fraction (defined under
1 B
k T condition) for various particle sizes and fluctuation magnitudes.
208
Fig. 7. 3 (B)
Elastic modulus is plotted against effective volume fraction (defined under 1 B
k T
condition) for various particle sizes and fluctuation magnitudes. The two horizontal lines
indicate the elastic modulus reported in experiment for 5nm and 7nm crosslinked
nanogels [].
209
Fig. 7. 4
The Maxwell viscosity is plotted against effective volume fraction for various fluctuation
amplitudes and three particle sizes.
210
CHAPTER 8
SUMMARY AND OUTLOOK
This dissertation has explored the key role of particle softness in two types of soft
condensed matter systems of importance in materials science and engineering: dense
suspensions of soft colloids and soft filler polymer nanocomposites. We attempted to do
so at a microscopic, predictive, single-length-scale to multi-length-scale level.
Although the experimental studies of soft colloids and nanoparticles have been
extensively carried out [1-8], unfortunately there has been little theoretical work done to
understand the role of particle softness on equilibrium structure and slow activated
dynamics. This statement has two aspects. The first concerns the way one models or
quantifies particle “softness”, i.e. at the particle-particle interaction level or at the surface
local interactions level. The second aspect deals with how one connects the interparticle
interaction to equilibrium structure, and later to thermodynamics and activated dynamics.
We have approached these aspects by combining the center-of-mass level pair potentials,
either well established based on connecting polymer coils to hard spheres [9] (many arm
stars interaction used in Chapter 3), or effectively mimicked with a one-length-scale
continuum mechanics model [4] (Hertzian model used in Chapter 4), or our new multi-
scale modeling method (our methodology developed in Chapter 5), with liquid state
theory. The computed structure is used as input to either nonlinear Langevin equation
theory to capture the slow activated dynamics (Chapter 2, 3, 4, 7), or the PRISM integral
equation approach to describe dispersion properties (Chapter 5, 6) in polymer
nanocomposites.
211
The first part of this thesis lies in the realm of glassy soft colloidal dynamics. We
applied the single-particle NLE theory [10] of activated dynamics to study the role of
particle softness in the context of dense suspension of either many arm stars [11] or
microgel-like Hertzian spheres [12,13]. Such model soft particle systems are key tools to
understand a wide variety of novel experimental phenomena that are labled as “soft
glasses”. We are able to qualitatively and quantitatively describe many such glassy
features, such as: the alpha relaxation time in the activated hopping regime is a rich
function of volume fraction and temperature, including approaching a maximum value at
ultra-high volume fraction due to a soft jamming crossover that signals local packing
disorder due to particle overlap. A kinetic arrest diagram is constructed, and its
qualitative features agree with the dynamic crossover (MCT) analog. The isothermal
dynamic fragility varies over a wide range, and soft particles are predicted to behave as
strong glasses, as experimentally observed [1].
Below the soft jamming threshold, the shear modulus roughly follows a power
law dependence on volume fraction with an apparent exponent that grows with repulsion
strength. The shear modulus varies inversely with the transient localization length under
all conditions studied. For the barrier and alpha relaxation time, to a first approximation
these local properties are controlled by a single coupling constant determined by local
fluid structure which quantifies the effective mean square force on a tagged particle. In
contrast to the behavior of hard spheres, an approximately linear relation between the
elastic modulus and activation barrier is predicted. This suggests a microscopic
foundation for the often observed, in both thermal liquids and soft microgel suspensions,
connections between a (relatively) short time/distance property and the long time/length
212
scale relaxation process [14]. For the Hertzian model, yield strains were found to be
relatively weakly dependent on volume fraction and single particle stiffness for
parameters relevant to typical microgel suspensions. Shear thinning is predicted to
commence at Peclet numbers far less than unity, a signature of relaxation via stress-
assisted activated barrier hopping. Power law thinning of the viscosity over many orders
of magnitude of shear rate is generically predicted but with an apparent exponent smaller
than unity. Approaching the soft jamming threshold, a nearly universal master flow curve
can be constructed at fixed repulsion strength which exhibits a power law form over
many intermediate orders of magnitude of reduced shear rate. The breadth and apparent
exponent in this regime systematically increases and decreases, respectively, as repulsion
strength and/or volume fraction grow.
Comparisons with available simulations and colloid experiments reveal at least
qualitative agreement. In particular, we found the non-monotonic change of the primary
peak of the pair correlation function, agreeing with both 2 dimensional microgel
experiment and 3 dimensional Hertzian sphere simulation [2], indicating the so-called
“soft jamming” transition. We also predict power law dependence of the elastic modulus
below the “soft jamming” volume fraction and the apparent power law exponent is
quantitatively close to experiment values. Above the “soft jamming” transition volume
fraction, a generally weaker dependence emerges, which agrees with recent experiments
on many arm stars (linear dependence) and microgels (weaker, some cases decreasing).
The scaling collapse of the mean barrier hopping time is achieved by using simulation
reported fitting parameters (Hertzian spheres). A general point is that by applying
microscopic theory for activated barrier hopping with an intuitive picture of center of
213
mass level interactions between two soft particles, we have made concrete, structure-
based connections with diverse glassy behaviors of dense suspensions of soft particles.
The second half of the dissertation studied the effective forces, equilibrium
structure and dispersion properties of soft filler polymer nanocomposites. Here, it is not
the interaction at particle scale, but rather the interaction and geometry constraints at the
surface corrugation scale comparable to monomer length scale, that largely determine
entropic depletion attraction between nanoparticles.
Multi-scale descriptions of polymer nanocomposites have been exceptionally
challenging to formulate, largely because exactly including all the internal and external
degrees of length scales exist in polymeric nanoparticles is too computationally massive
and conceptually difficult for surface morphology. Early attempts to understand the
thermodynamics of such nanocomposites relied on viewing the nanogel as a hard sphere,
and the entropically unfavorable dispersion is offset by an enthalpy gain due to an
increase in molecular contacts at dispersed nanoparticle surfaces as compared with the
surfaces of phase-separated nanoparticles [7]. But this idea ignores the fact that the
chemistry is matched, and the particle surface is corrugated and fluctuating on a scale
roughly of order the monomer size. As such, a multi-scale modling that explicitly studies
the role of surface morphology is desirable.
The core of our theoretical advances combined the small-scale MC simulation
with PRISM theory that accounts for the surfaces of such nanoparticles are corrugated
and fluctuating. This allowed the construction of the first minimalist, but representative,
multi-scale model that embeds the corrugation and fluctuation at the center-of-mass level
interparticle potentials: a soft repulsive interaction. However, whereas the first part of the
214
thesis (Chapter 4) uses one-length-scale potential (obtained, for example, from continuum
contact mechanics), this multi-scale modeling method allows for prediction of the particle
size dependence of all structural and dynamical (briefly discussed in Chapter 7) quantities.
Simple mappings are proposed between the rough HS filler (multi-scale) and smooth HS
filler (one scale) PNCs, which enabled a connection between local surface interaction to
three physical regimes in terms of bridging phenomena predicted by prior PRISM studies
[15]. This approach surely provides a better microscopic description of real nanogel
fillers.
Our results provide a physical basis for the unexpected ability to disperse
chemically matched crosslinked polystyrene in linear polystyrene melt. The surface
corrugation in the frozen surface regime results in a favorable entropic driving force for
mixing which competes with unfavorable depletion, resulting in a major enhancement of
nanoparticle dispersion, including a much larger spinodal solubility limit. Smooth hard
sphere behavior is recovered when the number of beads are significantly large (N=282)
and the relative corrugation size significantly reduced (from 20% to 10% of particle size).
When surface fluctuation exists, the dependence of solubility limit on fluctuation
magnitude is somewhat subtle, due to the relevance of multiple length scales. We find
incoherent surface fluctuation suppresses dispersion for all particles sizes (or surface
curvature) studied, and this effect is most pronounced when the monomer size is smaller
than corrugation size. The (presumably) less realistic extreme coherent fluctuation model
shows dramatically enhanced dependence on fluctuation magnitude at large fluctuation
magnitude and less miscibility. When interfacial cohesion exists, we have developed a
model for local bead-monomer level attraction, which explicitly connects to particle-
215
monomer level attraction through potential mapping strategies. By varying the interfacial
attraction strength, one can still observe all three regimes reported in prior PRISM studies
of smooth HS filler PNCs. Interestingly, the steric stabilization regime is not sensitive to
surface fluctuation magnitude, while the contact clustering and strong bridging regimes
are very sensitive to surface fluctuation magnitude. Surface fluctuation smears out
surface corrugation and reduces interfacial cohesion, leading to stronger depletion
attraction and more negative second virial coefficients.
With the advance of our multi-scale modeling of soft particles to treat two length
scales, we use the effective interaction to revisit the problem of slow activated dynamics
and elasticity of soft colloidal dense suspension in the context of the crosslinked nanogels,
focusing on the particle size dependence. Our shear elastic modulus calculations are in
qualitatively agreement with a recent experiment [8], in contrast with the well-known
smooth HS result. The elastic modulus grows with increasing particle size, while for
smooth HS suspensions the modulus is inversely proportional to the particle size cubed in
three dimensions. We plan to combine this advance with mechanical response theories to
describe the shear rheology of soft glasses from our multi-scale perspective. In particular,
the shear viscosity shows an even more pronounced dependence on particle size, which is
a direct result from the combination of the enhanced dependence on particle size for
elastic modulus and alpha relaxation time. This and other questions provide directions for
future research.
Our perspective in applying the multi-scale modeling developed in Chapter 5 to
melt of crosslinked nanogels is that although we still rely on center of mass level
interparticle potentials, the multi-scale modeling enables us to explicitly study the role of
216
particle size. The hope is that in the context of a two-length-scale center of mass potential,
one can capture information about the whole crosslinked network with a coarse-grained
description of hard spheres with corrugated beads that fluctuate incoherently. It is in this
sense we expect the explicit studies on monomer scale interaction will make us one step
closer to the real world nanogels.
Concerning possible future work, we note that single particle Brownian dynamics
simulation solution of Eq.(2.6) have previously been successfully employed to establish
the full trajectory level predictions of the nonlinear stochastic Langevin equation theory
of activated hopping dynamics in glassy hard sphere suspensions and fluids [16]. This
method can be used to explore the role of particle softness on the transport properties,
incoherent dynamic structure factor and dynamic heterogeneities of systems composed of
soft particles. Comparison of the results with experiments on many-arm stars and
crosslinked microgels could be performed, e.g. self-diffusion in a suspension of multi-
arm star 1,4-polybutadiene (267 arms) with an arm molecular weight of 4490 g/mol [17].
The experiment finds star self diffusive coefficients strongly decrease (but in a non-hard
sphere colloid manner) as the particle concentration increases, which is linked to the
strong increase of the viscosity [17]. The computed dynamical mean square displacement
can also be compared to confocal measurements on repulsive microgels [18].
The results and advances discussed in Chapter 2, 3, 4, 7 set the stage for
investigating more complicated situations for dense soft colloidal suspensions beyond the
two basic system parameters, volume fraction and particle softness. For example, one can
add short-range attractions of variable strength, or equivalently reduced temperature. For
hard spheres, this problem has been addressed based on MCT and NLE theories [10],
217
simulations [19] and experiments [20]. Many interesting behaviors have been found, such
as the emergence of gels and attractive glasses, a reentrant glass-fluid-gel transition,
nonmonotonic diffusivity as a function of attraction strength, and two-step yielding [21].
Different from hard spheres, now another type of complexity exists. Attractions exist on
the length scale of surface corrugation due to the polymeric nature of such soft particles.
We know that for attractive spheres, high volume fraction is no longer a necessary
condition to trap particles. As long as the attraction is strong enough, particles will be
transiently localized due to bonding to neighboring particles. However, whether this
physical bonding occurs on the particle length scale or on the surface corrugation length
scale perhaps largely determines the strength of such bonding and thus may strongly
modify the dynamic phase diagram.
Obtaining numerical converged solutions of PRISM theory for finite filler volume
fractions is a challenge. Finding a better algorithm is crucial for wider applications of the
PNC integral equation theory. If improved numerical methods work, then one can address
many body effects on miscibility and calculate the scattering structure factors that can be
compared with X-ray and neutron scattering data, and the phase diagrams can be
constructed.
The investigation of filler surface corrugation and fluctuation of PNCs systems
has already provided some guidelines to produce various desirable and undesirable
equilibrium structural behaviors. Since the dynamics is largely dependent on structural
input, we believe multi-scale modeling method can further provide guidance on the
viscoelasticity of polymer nanocomposites, which is of great importance to industry.
With stronger computational power, we can construct models that include more detailed
218
monomer level information. A possibly fruitful approach is to obtain effective
interactions based on our multi-scale modeling method, and determine mixture structure,
and input it to dynamical theories.
References
[1] J.Mattson, H.M.Wyss, A.Fernando-Nieves, K.Miyazaki, Z.Hu, D.R.Reichman and
D.A.Weitz, Nature, 462, 83 (2009).
[2] Z.Zhang, N.Xu, D.Chen, P.Yunker, A.M.Alsayed, K.B.Aptowicz, P.Habdas, A.J.Liu,
S.R.Nagel and A.G.Yodh, Nature, 459, 83 (2009).
[3] M.Cloitre, R.Borrega, F.Monti and L.Leibler, Phys.Rev.Lett., 90,068303 (2003).
[4]J.R.Seth, M.Cloitre and R.T.Bonnecaze,J.Rheology, 50, 353 (2006); 52, 1241 (2008).
[5] F.Ozon, G.Petekidis and D.Vlassopoulos, Ind. Eng. Chem. Res.,45, 6946 (2006);
M.E.Helgeson, N.J.Wagner and D.Vlassopoulos, J.Rheology, 51, 297 (2007).
[6] E.H.Purnomo, D.van den Ende, S.A.Vanapili and F.Mugele, Phys.Rev.Lett. 101,
23801 (2008).
[7] M.E.Mackay, A.Tuteja, P.M.Duxbury, C.J.Hawker, B. Van Horn and Z.H.Guan,
Science, 311, 1740 (2006).
[8] A,TUTEJA, M.E.MACKAY, C. J. HAWKER, B VAN HORN, D. L. HO, Journal of
Polymer Science: Part B: Polymer Physics, 2006
[9] Likos C.N., Phys.Rep., 348 (2001), 267; Jusufi A., Watzlawek M. and Lowen H.,
Macromolecules, 32, (1999), 4470.
[10] Schweizer K.S. and Saltzman E.J., J.Chem.Phys.,119 (2003), 1181; Schweizer K.S.,
J.Chem.Phys.,123(2005),244501;SchweizerK.S.,Curr.Opin.Coll.Inter.Sci.,12(2007), 297.
219
[11] J. Yang and K. S. Schweizer, Europhys. Lett., 2010, 90, 66001.
[12] J. Yang and K. S. Schweizer, J. Chem. Phys. 134, 204908 (2011).
[13] J. Yang and K. S. Schweizer, J. Chem. Phys. 134, 204909 (2011).
[14] J. Dyre, Rev. Mod. Phys., 78, 953 (2006).
[15] J. B. Hooper and K. S. Schweizer, Macromolecules 39, 5133 (2006).
[16] E.J.Saltzman and K.S.Schweizer, J. Chem. Phys. 125, 044509 (2006)
[17] F.Ozon, G.Petekidis and D.Vlassopoulos, Ind. Eng. Chem. Res.,45, 6946 (2006)
[18] M.Cloitre, R.Borrega, F.Monti and L.Leibler, Phys.Rev.Lett., 90,068303 (2003)
[19] E. Zaccarelli and W. C. K. Poon, Proc. Natl. Acad. Sci. 106, 15203 (2009).
[20] K. N. Pham, A. M. Puertas, J. Bergenholtz, S. U. Egelhaaf, A. Moussaïd, P. N.
Pusey, A. B. Schofield, M. E. Cates, M .Fuchs, and W. C. K. Poon, Science 296, 104
(2002).
[21] K. N. Pham, G. Petekidis, D. Vlassopoulos, S. U. Egelhaaf, W. C. K. Poon, and P. N.
Pusey, J. Rheology 52, 649 (2008)